From 3d77f96eba1301b08a6eba38f1e6bb6c8b776061 Mon Sep 17 00:00:00 2001 From: Neil <4138956+nwparker@users.noreply.github.com> Date: Wed, 11 Nov 2026 18:39:16 -0800 Subject: [PATCH 01/13] fix(artifacts): trust orca.dev alongside onorca.dev (#41324) --- src/main/artifacts/artifact-cloud-config.test.ts | 14 ++++++++++++++ src/main/artifacts/artifact-cloud-config.ts | 6 ++++-- 2 files changed, 18 insertions(+), 2 deletions(-) diff --git a/src/main/artifacts/artifact-cloud-config.test.ts b/src/main/artifacts/artifact-cloud-config.test.ts index 7d59e841818..52c6752fcbc 100644 --- a/src/main/artifacts/artifact-cloud-config.test.ts +++ b/src/main/artifacts/artifact-cloud-config.test.ts @@ -25,6 +25,20 @@ describe('resolveArtifactCloudApiUrl', () => { ) }) + it('allows both first-party domain families without trusting similar hostnames', () => { + for (const origin of ['https://share.onorca.dev', 'https://share.orca.dev']) { + expect(resolveArtifactCloudApiUrl(origin, {}, true)).toBe(origin) + } + for (const origin of [ + 'https://share.orca.dev.example.com', + 'https://notorca.dev', + 'https://notonorca.dev', + 'https://share.onorca.dev.example.com' + ]) { + expect(() => resolveArtifactCloudApiUrl(origin, {}, true)).toThrow() + } + }) + it('allows auth token overrides only in non-production development builds', () => { expect(allowsArtifactCloudAuthOverride({}, false)).toBe(true) expect(allowsArtifactCloudAuthOverride({ NODE_ENV: 'production' }, false)).toBe(false) diff --git a/src/main/artifacts/artifact-cloud-config.ts b/src/main/artifacts/artifact-cloud-config.ts index 6ff5af72938..d8b4206eba6 100644 --- a/src/main/artifacts/artifact-cloud-config.ts +++ b/src/main/artifacts/artifact-cloud-config.ts @@ -18,12 +18,14 @@ export function resolveArtifactCloudApiUrl( const candidate = override?.trim() || env.ORCA_ARTIFACTS_API_URL?.trim() const url = new URL(candidate || PRODUCTION_ARTIFACTS_API_URL) const loopback = ['127.0.0.1', 'localhost', '[::1]'].includes(url.hostname) - const firstParty = url.hostname === 'onorca.dev' || url.hostname.endsWith('.onorca.dev') + const firstParty = ['onorca.dev', 'orca.dev'].some( + (domain) => url.hostname === domain || url.hostname.endsWith(`.${domain}`) + ) if (url.protocol !== 'https:' && !(url.protocol === 'http:' && loopback && !packaged)) { throw new Error('Artifact API URLs must use HTTPS; local development may use loopback HTTP.') } if (!firstParty && !loopback) { - throw new Error('Artifact API URLs must use an onorca.dev or loopback host.') + throw new Error('Artifact API URLs must use an onorca.dev, orca.dev, or loopback host.') } if (url.username || url.password || url.search || url.hash || url.pathname !== '/') { throw new Error('Artifact API URL must be an origin without credentials, paths, or parameters.') From 1d3bf2d6a9009e8afc3a2179903a7f5bde6b8dec Mon Sep 17 00:00:00 2001 From: Kelvin Amoaba <97001695+AmoabaKelvin@users.noreply.github.com> Date: Thu, 12 Nov 2026 02:41:44 +0000 Subject: [PATCH 02/13] fix(mobile): send chat with Cmd+Enter in desktop Mac browsers (#41330) --- mobile/src/platform/host-os.web.test.ts | 2 +- mobile/src/platform/host-os.web.ts | 7 +++++-- mobile/src/platform/primary-modifier-key.ts | 7 +++++++ .../MobileNativeChatComposer.keyboard.web.test.tsx | 12 ++++++++++++ .../composer/MobileNativeChatComposerField.web.tsx | 6 ++++-- ...use-mobile-native-chat-steer-chord-preview.web.ts | 5 ++--- 6 files changed, 31 insertions(+), 8 deletions(-) create mode 100644 mobile/src/platform/primary-modifier-key.ts diff --git a/mobile/src/platform/host-os.web.test.ts b/mobile/src/platform/host-os.web.test.ts index fcc3d446f36..7673afefbbb 100644 --- a/mobile/src/platform/host-os.web.test.ts +++ b/mobile/src/platform/host-os.web.test.ts @@ -33,6 +33,6 @@ describe("the page's host OS", () => { agent('Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/605.1.15', 'MacIntel', 5) expect(hostOs()).toBe('ios') agent('Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/605.1.15', 'MacIntel', 0) - expect(hostOs()).toBe('web') + expect(hostOs()).toBe('macos') }) }) diff --git a/mobile/src/platform/host-os.web.ts b/mobile/src/platform/host-os.web.ts index 2f83fb2c7f0..1fa67c726d6 100644 --- a/mobile/src/platform/host-os.web.ts +++ b/mobile/src/platform/host-os.web.ts @@ -2,7 +2,7 @@ import type { HostOs } from './host-os' export type { HostOs } from './host-os' -// Use the phone OS for keyboard/inset arithmetic: Platform.OS is web here; iPadOS needs the Mac-plus-touch heuristic. +// Platform.OS is web here: read the phone OS for keyboard/inset arithmetic (iPadOS needs the Mac-plus-touch heuristic) and a desktop Mac for Cmd chords. export function hostOs(): HostOs { const agent = globalThis.navigator?.userAgent ?? '' if (/iP(ad|hone|od)/.test(agent)) { @@ -11,5 +11,8 @@ export function hostOs(): HostOs { if (globalThis.navigator?.platform === 'MacIntel' && globalThis.navigator.maxTouchPoints > 1) { return 'ios' } - return agent.includes('Android') ? 'android' : 'web' + if (agent.includes('Android')) { + return 'android' + } + return agent.includes('Macintosh') ? 'macos' : 'web' } diff --git a/mobile/src/platform/primary-modifier-key.ts b/mobile/src/platform/primary-modifier-key.ts new file mode 100644 index 00000000000..769305b0da6 --- /dev/null +++ b/mobile/src/platform/primary-modifier-key.ts @@ -0,0 +1,7 @@ +import { hostOs } from './host-os' + +/** The submit chord's modifier: Cmd on an Apple keyboard (iPhone, iPad, Mac), Ctrl elsewhere. */ +export function primaryModifierKey(): 'Meta' | 'Control' { + const os = hostOs() + return os === 'ios' || os === 'macos' ? 'Meta' : 'Control' +} diff --git a/mobile/src/session/native-chat/composer/MobileNativeChatComposer.keyboard.web.test.tsx b/mobile/src/session/native-chat/composer/MobileNativeChatComposer.keyboard.web.test.tsx index 3facb1e14f7..8a1960ed425 100644 --- a/mobile/src/session/native-chat/composer/MobileNativeChatComposer.keyboard.web.test.tsx +++ b/mobile/src/session/native-chat/composer/MobileNativeChatComposer.keyboard.web.test.tsx @@ -188,6 +188,18 @@ describe('mobile composer modifier submit in the paired web page', () => { expect(field().value).toBe(' hello ') }) + it('sends with Cmd+Enter, not Ctrl+Enter, in a desktop Mac browser', async () => { + vi.spyOn(navigator, 'userAgent', 'get').mockReturnValue( + 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/142.0.0.0 Safari/537.36' + ) + vi.spyOn(navigator, 'platform', 'get').mockReturnValue('MacIntel') + await render() + expect((await key({ ctrlKey: true })).defaultPrevented).toBe(false) + expect(onSend).not.toHaveBeenCalled() + expect((await key({ metaKey: true })).defaultPrevented).toBe(true) + expect(onSend).toHaveBeenCalledTimes(1) + }) + it.each([ { value: '@rea', expected: '@README.md ', filePaths: ['README.md', 'src/read.ts'] }, { value: '/comp', expected: '/compact ', filePaths: [] } diff --git a/mobile/src/session/native-chat/composer/MobileNativeChatComposerField.web.tsx b/mobile/src/session/native-chat/composer/MobileNativeChatComposerField.web.tsx index 9e535488227..0cf471bfc88 100644 --- a/mobile/src/session/native-chat/composer/MobileNativeChatComposerField.web.tsx +++ b/mobile/src/session/native-chat/composer/MobileNativeChatComposerField.web.tsx @@ -8,7 +8,7 @@ import { type TextInputInstance } from 'react-native' import { useKeyboardOcclusion } from '../../../platform/keyboard-occlusion' -import { hostOs } from '../../../platform/host-os' +import { primaryModifierKey } from '../../../platform/primary-modifier-key' import { useTextInputContentHeight } from '../../../platform/text-input-content-height' import { useTextInputSelectionReveal } from '../../../platform/text-input-selection-reveal.web' import { colors } from '../../../theme/mobile-theme' @@ -120,7 +120,9 @@ export function MobileNativeChatComposerField({ key.keyCode === 229 || key.shiftKey || key.altKey || - !(hostOs() === 'ios' ? key.metaKey && !key.ctrlKey : key.ctrlKey && !key.metaKey) + !(primaryModifierKey() === 'Meta' + ? key.metaKey && !key.ctrlKey + : key.ctrlKey && !key.metaKey) ) { return } diff --git a/mobile/src/session/native-chat/composer/use-mobile-native-chat-steer-chord-preview.web.ts b/mobile/src/session/native-chat/composer/use-mobile-native-chat-steer-chord-preview.web.ts index ae72d6d9c32..57fb8783602 100644 --- a/mobile/src/session/native-chat/composer/use-mobile-native-chat-steer-chord-preview.web.ts +++ b/mobile/src/session/native-chat/composer/use-mobile-native-chat-steer-chord-preview.web.ts @@ -1,5 +1,5 @@ import { useEffect, useState } from 'react' -import { hostOs } from '../../../platform/host-os' +import { primaryModifierKey } from '../../../platform/primary-modifier-key' import { createNativeChatSteerChordPreviewHook } from '../../../../../src/shared/native-chat/input/use-native-chat-steer-chord-preview' import type { MobileNativeChatSteerChordPreviewInput } from './use-mobile-native-chat-steer-chord-preview' @@ -36,8 +36,7 @@ export function useMobileNativeChatSteerChordPreview({ steerHead && (() => steerHead.run(alternateSteers)), sendAlternate, pickerOpen, - // The field's own submit chord: Cmd on an iPhone or iPad keyboard, Ctrl elsewhere. - hostOs() === 'ios' ? 'Meta' : 'Control' + primaryModifierKey() ) const lit = chord.preview === 'steer-queued' const light = steerHead?.light From 9f32c6d06dd692824b888a1eb6b089cee4fb00b7 Mon Sep 17 00:00:00 2001 From: OrcaWin Date: Wed, 11 Nov 2026 18:44:33 -0800 Subject: [PATCH 03/13] test: remove worker approval smoke and duplicate refusal cases (#41331) Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com> --- src/main/ipc/repos-nested-import.test.ts | 15 ---- src/main/ipc/repos-nested-scan.test.ts | 10 --- .../crew/crew-worker-approvals.test.ts | 29 ------- .../terminal-worker-approval-holds.test.ts | 64 +------------- .../worker-approval-stream-wakes.test.ts | 36 -------- .../NativeChatMessageList.timed-rows.test.tsx | 62 ------------- .../NativeChatWorkerApprovals.test.tsx | 86 ------------------- .../native-chat-worker-approval-feed.test.ts | 9 -- 8 files changed, 2 insertions(+), 309 deletions(-) delete mode 100644 src/main/runtime/workspaces/persisted-surfaces/worker-approval-stream-wakes.test.ts delete mode 100644 src/renderer/src/components/native-chat/conversation/slots/NativeChatMessageList.timed-rows.test.tsx delete mode 100644 src/renderer/src/components/native-chat/worker-approvals/NativeChatWorkerApprovals.test.tsx diff --git a/src/main/ipc/repos-nested-import.test.ts b/src/main/ipc/repos-nested-import.test.ts index 2ffd090598b..63cada51bf9 100644 --- a/src/main/ipc/repos-nested-import.test.ts +++ b/src/main/ipc/repos-nested-import.test.ts @@ -148,19 +148,4 @@ describe('projectGroups IPC validation', () => { expect(mockStore.addRepo).not.toHaveBeenCalled() }) }) - - it('refuses nested imports on an SSH target and never inspects the path locally', async () => { - await expect( - handlers.get('projectGroups:importNested')!(null, { - parentPath: '/srv/platform', - groupName: 'Platform', - projectPaths: ['/srv/platform/api'], - connectionId: 'conn-1', - mode: 'group' - }) - ).rejects.toThrow('ssh_connection_unavailable') - - expect(mockStore.addRepo).not.toHaveBeenCalled() - expect(isGitRepo).not.toHaveBeenCalled() - }) }) diff --git a/src/main/ipc/repos-nested-scan.test.ts b/src/main/ipc/repos-nested-scan.test.ts index d454c66ffaf..a11875663b3 100644 --- a/src/main/ipc/repos-nested-scan.test.ts +++ b/src/main/ipc/repos-nested-scan.test.ts @@ -84,16 +84,6 @@ describe('projectGroups IPC validation', () => { registerRepoHandlers(mockWindow as never, mockStore as never, {} as never) }) - it('refuses SSH nested scans without reading the remote or the local filesystem', async () => { - await expect( - handlers.get('projectGroups:scanNested')!(null, { - path: '/srv/platform', - connectionId: 'conn-1' - }) - ).rejects.toThrow('ssh_connection_unavailable') - expect(isGitRepo).not.toHaveBeenCalled() - }) - it('returns partial local scan results after cancellation', async () => { vi.mocked(isGitRepo).mockReturnValue(false) const root = await mkdtemp(join(tmpdir(), 'orca-nested-local-cancel-')) diff --git a/src/main/runtime/orchestration/crew/crew-worker-approvals.test.ts b/src/main/runtime/orchestration/crew/crew-worker-approvals.test.ts index 5ee01ca8dc0..9c876321f34 100644 --- a/src/main/runtime/orchestration/crew/crew-worker-approvals.test.ts +++ b/src/main/runtime/orchestration/crew/crew-worker-approvals.test.ts @@ -84,7 +84,6 @@ it('keeps showing an open approval behind more closed ones than the row bound', const [row] = buildCrewSnapshot('term_coord', deps).sections.managing const ids = row.approvals?.map(({ approval }) => approval.approvalId) - expect(ids).toHaveLength(100) expect(ids?.[0]).toBe('apr_still_open') expect(ids?.at(-1)).toBe('apr_149') }) @@ -124,31 +123,3 @@ it('closes an open approval once its Dispatch stops running, at the time it ende ]) expect(row.approvals?.[0].approval.expiresAt).toBeUndefined() }) - -it('reads a close no one named, after its Dispatch ended, as the worker ending', () => { - const { db, deps, run, assign } = fixture() - const dispatch = assign('term_worker') - db.insertMessage({ - id: approvalReportId(dispatch.id, 'apr_unnamed', 'closed'), - from: 'term_worker', - to: `run:${run.id}`, - subject: 'apr_unnamed', - type: 'status', - runId: run.id, - payload: JSON.stringify({ - approval: { - v: 1, - approvalId: 'apr_unnamed', - state: 'answered', - worker: { handle: 'term_worker', dispatchId: dispatch.id, taskId: dispatch.task_id }, - summary: { title: 'Allow Bash?', detail: null, reason: null }, - allowedActions: [], - outcome: { by: 'unknown', decision: null, reason: null, reasonDelivered: false } - } - }) - }) - db.completeDispatch(dispatch.id) - - const [row] = buildCrewSnapshot('term_coord', deps).sections.managing - expect(row.approvals?.map(({ approval }) => approval.outcome?.by)).toEqual(['worker-ended']) -}) diff --git a/src/main/runtime/orchestration/workers/approvals/terminal-worker-approval-holds.test.ts b/src/main/runtime/orchestration/workers/approvals/terminal-worker-approval-holds.test.ts index 9b61f0b8767..16fc46dbd99 100644 --- a/src/main/runtime/orchestration/workers/approvals/terminal-worker-approval-holds.test.ts +++ b/src/main/runtime/orchestration/workers/approvals/terminal-worker-approval-holds.test.ts @@ -43,16 +43,6 @@ const USER_NO: ClaudeToolResult = { isError: true, text: "The user doesn't want to proceed with this tool use. The tool use was rejected." } -const AUTO_BLOCK: ClaudeToolResult = { - isError: true, - text: 'Permission for this command was denied by a built-in Claude Code safety check, not by the user.' -} -// An allowed command that then fails (probe 2026-11-11, the owner's live case). -const FAILED: ClaudeToolResult = { - isError: true, - text: "Exit code 1\n(eval):14: parse error near `\\n'" -} -const RAN: ClaudeToolResult = { isError: false, text: '59' } /** What the transcript holds: the hook decision Claude took, then the tool's result. */ const record = ( result: ClaudeToolResult | null, @@ -147,14 +137,11 @@ function hook( } } -function postToolUse( - toolUseId: string, - hookEventName: 'PostToolUse' | 'PostToolUseFailure' = 'PostToolUse' -): AgentHookEventPayload { +function postToolUse(toolUseId: string): AgentHookEventPayload { return { paneKey: PANE, connectionId: null, - hookEventName, + hookEventName: 'PostToolUse', toolUseId, payload: { state: 'working', prompt: '', agentType: 'claude' } } @@ -232,23 +219,6 @@ describe('terminal worker approval holds', () => { expect(ctx.closedWith()[0]).toMatchObject({ state: 'answered', by: 'unknown' }) }) - it.each([ - ['allowed, and the command succeeds', record(RAN, 'allow')], - ['allowed, and the command fails', record(FAILED, 'allow')], - ['allowed, and the command is still running', record(null, 'allow')], - ['allowed by a Claude that records no hook decision, and the command fails', record(FAILED)] - ])("takes Orca's allow when %s", async (_, written) => { - const ctx = setup() - const asked = hook('tool-1', () => { - ctx.screen.lines = AFTER - ctx.screen.record = written - }) - const approvalId = await opened(ctx, asked) - - expect(await answer(ctx, approvalId, 'allow')).toMatchObject({ ok: true }) - expect(asked.replies).toEqual([{ behavior: 'allow' }]) - }) - it("doesn't take Orca's allow when the person's No won the race", async () => { const ctx = setup() const approvalId = await opened( @@ -271,36 +241,6 @@ describe('terminal worker approval holds', () => { ]) }) - it("names a Yes in the terminal whose command failed as allowed in the worker's terminal", async () => { - const ctx = setup() - await opened(ctx, hook('tool-1')) - ctx.screen.lines = AFTER - ctx.holds.observe(postToolUse('tool-1', 'PostToolUseFailure')) - - expect(ctx.closedWith()).toEqual([ - { - state: 'answered', - by: 'worker-terminal', - decision: 'allow', - reason: null, - reasonDelivered: false - } - ]) - }) - - it("names Claude's 2:00 auto-block as an expiry", async () => { - const ctx = setup() - await closedInTerminal(ctx, AUTO_BLOCK) - - expect(ctx.closedWith()[1]).toEqual({ - state: 'expired', - by: 'auto', - decision: 'deny', - reason: null, - reasonDelivered: false - }) - }) - it("tells the coordinator to send the user to the terminal when the hook is gone but the dialog isn't", async () => { const ctx = setup() const asked = hook('tool-1') diff --git a/src/main/runtime/workspaces/persisted-surfaces/worker-approval-stream-wakes.test.ts b/src/main/runtime/workspaces/persisted-surfaces/worker-approval-stream-wakes.test.ts deleted file mode 100644 index c2218aac4a7..00000000000 --- a/src/main/runtime/workspaces/persisted-surfaces/worker-approval-stream-wakes.test.ts +++ /dev/null @@ -1,36 +0,0 @@ -import { expect, it, vi } from 'vitest' -import { OrcaRuntimeService } from '../../orca-runtime' -import { OrchestrationDb } from '../../orchestration/db/orchestration-db' -import { isOrcaSessionId } from '../../../../shared/orca-session-address' - -const CHAT = 'claude_11111111_1111_4111_8111_111111111111' - -it("reads a chat's approvals on its own Run's changes, never on terminal identity notices", async () => { - if (!isOrcaSessionId(CHAT)) { - throw new Error('fixture id') - } - const runtime = new OrcaRuntimeService() - const db = new OrchestrationDb(':memory:') - runtime.setOrchestrationDb(db) - const runs = vi.spyOn(db, 'runsBoundToCoordinator') - const stop = runtime.subscribeOrchestrationWorkerApprovals(`orca_session_id:${CHAT}`, () => {}) - const settle = () => new Promise((resolve) => setImmediate(resolve)) - const reads = runs.mock.calls.length - - // What every agent status or tab title change sends: terminal identity's default topology notice. - for (let i = 0; i < 50; i++) { - runtime.crewChanges.notify() - await settle() - } - expect(runs.mock.calls.length - reads).toBe(0) - - db.createRun({ - objective: 'Coordinate', - coordinatorHandle: null, - coordinatorPaneKey: null, - coordinatorOrcaSessionId: CHAT - }) - await settle() - expect(runs.mock.calls.length - reads).toBe(1) - stop() -}) diff --git a/src/renderer/src/components/native-chat/conversation/slots/NativeChatMessageList.timed-rows.test.tsx b/src/renderer/src/components/native-chat/conversation/slots/NativeChatMessageList.timed-rows.test.tsx deleted file mode 100644 index 1091db5d524..00000000000 --- a/src/renderer/src/components/native-chat/conversation/slots/NativeChatMessageList.timed-rows.test.tsx +++ /dev/null @@ -1,62 +0,0 @@ -// @vitest-environment happy-dom - -import '@testing-library/jest-dom/vitest' - -import { cleanup, screen } from '@testing-library/react' -import { afterEach, beforeEach, expect, it, vi } from 'vitest' -import { render } from '../../messages/actions/native-chat-tooltip-test-harness' -import { NativeChatMessageList } from '../../NativeChatMessageList' -import { - layout, - marker, - session, - stubLayout, - stubResizeObserver -} from '../../viewport/virtualization/native-chat-windowing-test-harness' - -// Pacing is not this test's subject. -vi.mock('../../markdown/use-native-chat-paced-text', async (importOriginal) => - (await import('../../markdown/native-chat-unpaced-text-fixture')).unpacedTextModule( - importOriginal - ) -) - -let restore: (() => void)[] = [] -beforeEach(() => { - restore = [stubLayout({ scrollGeometry: true, offsetChain: true }), stubResizeObserver()] -}) -afterEach(() => { - cleanup() - for (const undo of restore.toReversed()) { - undo() - } - layout.measuredRowHeights = [] -}) - -it('draws a worker approval at its own time, with newer replies after it', () => { - // marker(i) is message `marker-i`, at time i + 1. - const messages = [0, 1, 2, 3].map(marker) - render( - Allow Bash? (open)

}, - { key: 'receipt', at: 1.5, node:

Allowed by you 3:11 PM

} - ]} - /> - ) - - const order = [ - 'marker-0', - 'Allowed by you 3:11 PM', - 'marker-1', - 'marker-2', - 'Allow Bash? (open)', - 'marker-3' - ].map((text) => screen.getByText(text)) - for (const [index, element] of order.slice(1).entries()) { - expect(order[index].compareDocumentPosition(element)).toBe(Node.DOCUMENT_POSITION_FOLLOWING) - } -}) diff --git a/src/renderer/src/components/native-chat/worker-approvals/NativeChatWorkerApprovals.test.tsx b/src/renderer/src/components/native-chat/worker-approvals/NativeChatWorkerApprovals.test.tsx deleted file mode 100644 index 208f91bceae..00000000000 --- a/src/renderer/src/components/native-chat/worker-approvals/NativeChatWorkerApprovals.test.tsx +++ /dev/null @@ -1,86 +0,0 @@ -// @vitest-environment happy-dom - -import '@testing-library/jest-dom/vitest' - -import { cleanup, fireEvent, render, screen, waitFor } from '@testing-library/react' -import { afterEach, expect, it, vi } from 'vitest' -import type { OrchestrationCrewRow } from '../../../../../shared/orchestration/orchestration-crew' - -const rpc = vi.hoisted(() => ({ call: vi.fn() })) -const toast = vi.hoisted(() => ({ success: vi.fn(), info: vi.fn(), error: vi.fn() })) -vi.mock('@/runtime/rpc/calls/runtime-rpc-call', () => ({ callRuntimeRpc: rpc.call })) -vi.mock('sonner', () => ({ toast })) - -const { RuntimeRpcCallError } = await import('@/runtime/rpc/calls/runtime-rpc-result') -const { NativeChatWorkerApprovalStack } = await import('./NativeChatWorkerApprovals') - -afterEach(() => { - cleanup() - vi.clearAllMocks() -}) - -const worker: OrchestrationCrewRow = { - partyKey: 'w', - role: 'worker', - label: 'try-claude', - agentKind: 'claude', - status: 'blocked' -} -const entry = { - approval: { - v: 1 as const, - approvalId: 'apr_1', - state: 'open' as const, - worker: { handle: 'term_w', dispatchId: 'ctx_1', taskId: 'task_1' }, - summary: { title: 'Allow Bash?', detail: 'rg -n x', reason: null }, - allowedActions: ['allow' as const, 'deny' as const] - }, - at: '2026-11-11T20:11:30.000Z', - askedAt: '2026-11-11T20:11:24.000Z', - worker -} - -const reject = (code: string) => - rpc.call.mockRejectedValueOnce( - new RuntimeRpcCallError({ - id: '1', - ok: false, - error: { code, message: code }, - _meta: { runtimeId: 'r' } - }) - ) - -function renderStack() { - render() -} - -it('waits on the host after a click: both buttons off, the clicked one working, no toast when it lands', async () => { - renderStack() - let land: (value: unknown) => void = () => {} - rpc.call.mockReturnValueOnce(new Promise((resolve) => (land = resolve))) - fireEvent.click(screen.getByRole('button', { name: 'Allow' })) - expect(screen.getByRole('button', { name: 'Allow' })).toBeDisabled() - expect(screen.getByRole('button', { name: 'Deny' })).toBeDisabled() - expect( - screen.getByRole('button', { name: 'Allow' }).querySelector('.animate-spin') - ).not.toBeNull() - expect(screen.getByRole('button', { name: 'Deny' }).querySelector('.animate-spin')).toBeNull() - land({}) - await waitFor(() => expect(screen.getByRole('button', { name: 'Allow' })).toBeEnabled()) - expect(toast.success).not.toHaveBeenCalled() - expect(toast.info).not.toHaveBeenCalled() -}) - -it('leaves the card to the host list when the answer was not applied', async () => { - renderStack() - reject('approval_not_open') - fireEvent.click(screen.getByRole('button', { name: 'Allow' })) - await waitFor(() => expect(toast.info).toHaveBeenCalledTimes(1)) - // Still open as far as this card knows: answerable again until the host's list says otherwise. - expect(screen.getByRole('button', { name: 'Allow' })).toBeEnabled() - - reject('internal_error') - fireEvent.click(screen.getByRole('button', { name: 'Deny' })) - await waitFor(() => expect(toast.error).toHaveBeenCalledTimes(1)) - expect(screen.getByRole('button', { name: 'Deny' })).toBeEnabled() -}) diff --git a/src/renderer/src/components/native-chat/worker-approvals/native-chat-worker-approval-feed.test.ts b/src/renderer/src/components/native-chat/worker-approvals/native-chat-worker-approval-feed.test.ts index 4b5df1e6238..2d2561d0b3b 100644 --- a/src/renderer/src/components/native-chat/worker-approvals/native-chat-worker-approval-feed.test.ts +++ b/src/renderer/src/components/native-chat/worker-approvals/native-chat-worker-approval-feed.test.ts @@ -82,12 +82,3 @@ it('keeps a prompt that closed while shown pinned as its receipt for a moment, t expect(pinned.result.current).toEqual([]) expect(receipts.result.current.map((e) => e.approval.approvalId)).toEqual(['apr_old', 'apr_1']) }) - -it('keeps a prompt only the worker terminal can answer pinned', async () => { - vi.useFakeTimers() - const pinned = renderHook(() => feed.useNativeChatPinnedWorkerApprovals('orca_session_id:other')) - await act(async () => {}) - send(approval('apr_t', 'answered', 'terminal-only')) - act(() => vi.advanceTimersByTime(60_000)) - expect(pinned.result.current.map((e) => e.approval.approvalId)).toEqual(['apr_t']) -}) From a0d34074ea367f92ecde4609a7581e4f96135bd7 Mon Sep 17 00:00:00 2001 From: Kelvin Amoaba <97001695+AmoabaKelvin@users.noreply.github.com> Date: Thu, 12 Nov 2026 03:03:05 +0000 Subject: [PATCH 04/13] fix(native-chat): send Claude images over 5 MB instead of refusing the message (#41333) Orca refused any image over 5 MiB before handing the message to Claude. That is the Anthropic API's per-image limit, but Claude Code shrinks every image it receives over stream-json (to 2000 px, then a ~500 KB JPEG) before the API sees it, so the check only blocked images Claude would have sent. The per-image cap now matches the 50 MiB attachment upload cap and the per-message total is 64 MiB; both only bound host memory. --- ...ured-dispatch-attachment-rejection.test.ts | 20 +++++++------- ...claude-structured-dispatch-content.test.ts | 26 +++++++++++++++++++ .../claude-structured-dispatch-content.ts | 7 ++--- 3 files changed, 41 insertions(+), 12 deletions(-) diff --git a/src/main/claude/structured/claude-structured-dispatch-attachment-rejection.test.ts b/src/main/claude/structured/claude-structured-dispatch-attachment-rejection.test.ts index 813559ff937..919b1dd685d 100644 --- a/src/main/claude/structured/claude-structured-dispatch-attachment-rejection.test.ts +++ b/src/main/claude/structured/claude-structured-dispatch-attachment-rejection.test.ts @@ -1,4 +1,4 @@ -import { mkdtemp, rm, writeFile } from 'node:fs/promises' +import { mkdtemp, rm, truncate, writeFile } from 'node:fs/promises' import { tmpdir } from 'node:os' import { join } from 'node:path' import { describe, expect, it, vi } from 'vitest' @@ -25,13 +25,14 @@ describe('Claude structured dispatch attachment rejections', () => { expect(session.connection.send).not.toHaveBeenCalled() }) - it('rejects local images whose aggregate size exceeds twenty MiB', async () => { + it('rejects local images whose aggregate size exceeds sixty-four MiB', async () => { const directory = await mkdtemp(join(tmpdir(), 'orca-claude-images-')) try { const paths = await Promise.all( - Array.from({ length: 5 }, async (_, index) => { + Array.from({ length: 2 }, async (_, index) => { const path = join(directory, `${index}.png`) - await writeFile(path, Buffer.alloc(5 * 1024 * 1024)) + await writeFile(path, '') + await truncate(path, 32 * 1024 * 1024 + 1) return path }) ) @@ -43,10 +44,10 @@ describe('Claude structured dispatch attachment rejections', () => { ).resolves.toEqual({ state: 'rejected', reason: - 'The images on this message add up to more than 20 MB, so the message was not sent.', + 'The images on this message add up to more than 64 MB, so the message was not sent.', rejection: { kind: 'attachmentInvalid', - attachment: { reason: 'totalTooLarge', limit: 20 * 1024 * 1024 } + attachment: { reason: 'totalTooLarge', limit: 64 * 1024 * 1024 } } }) expect(session.connection.send).not.toHaveBeenCalled() @@ -59,7 +60,8 @@ describe('Claude structured dispatch attachment rejections', () => { const directory = await mkdtemp(join(tmpdir(), 'orca-claude-image-')) try { const path = join(directory, 'oversized.png') - await writeFile(path, Buffer.alloc(5 * 1024 * 1024 + 1)) + await writeFile(path, '') + await truncate(path, 50 * 1024 * 1024 + 1) const session = sessionFor() const body = userMessage([{ type: 'image-ref', path }]) @@ -67,10 +69,10 @@ describe('Claude structured dispatch attachment rejections', () => { dispatchClaudeTurn(session, { clientMessageId: 'client-1', body }) ).resolves.toEqual({ state: 'rejected', - reason: 'An image on this message is larger than 5 MB, so the message was not sent.', + reason: 'An image on this message is larger than 50 MB, so the message was not sent.', rejection: { kind: 'attachmentInvalid', - attachment: { reason: 'tooLarge', limit: 5 * 1024 * 1024 } + attachment: { reason: 'tooLarge', limit: 50 * 1024 * 1024 } } }) expect(session.connection.send).not.toHaveBeenCalled() diff --git a/src/main/claude/structured/claude-structured-dispatch-content.test.ts b/src/main/claude/structured/claude-structured-dispatch-content.test.ts index d6d19c1e42b..8853e02215c 100644 --- a/src/main/claude/structured/claude-structured-dispatch-content.test.ts +++ b/src/main/claude/structured/claude-structured-dispatch-content.test.ts @@ -1,3 +1,6 @@ +import { mkdtemp, rm, writeFile } from 'node:fs/promises' +import { tmpdir } from 'node:os' +import { join } from 'node:path' import { describe, expect, it } from 'vitest' import type { AgentJournalMessageItem } from '../../../shared/agent-session/journal/agent-session-journal-types' import { claudeDispatchMessageContent } from './claude-structured-dispatch-content' @@ -19,4 +22,27 @@ describe('claudeDispatchMessageContent', () => { { type: 'text', text: '/goal ship the parser' } ]) }) + + it('sends a local image over the API per-image limit for Claude to shrink', async () => { + const directory = await mkdtemp(join(tmpdir(), 'orca-claude-image-')) + try { + const path = join(directory, 'retina-screenshot.png') + await writeFile(path, Buffer.alloc(6 * 1024 * 1024, 1)) + + const content = await claudeDispatchMessageContent(userMessage([{ type: 'image-ref', path }])) + + expect(content).toEqual([ + { + type: 'image', + source: { + type: 'base64', + media_type: 'image/png', + data: Buffer.alloc(6 * 1024 * 1024, 1).toString('base64') + } + } + ]) + } finally { + await rm(directory, { recursive: true, force: true }) + } + }) }) diff --git a/src/main/claude/structured/claude-structured-dispatch-content.ts b/src/main/claude/structured/claude-structured-dispatch-content.ts index 3d3f691bbef..43f8c0dc60f 100644 --- a/src/main/claude/structured/claude-structured-dispatch-content.ts +++ b/src/main/claude/structured/claude-structured-dispatch-content.ts @@ -13,6 +13,7 @@ import { type AgentJournalDispatchRejection } from '../../../shared/agent-session/failures/agent-session-failure-words' import { TUI_AGENT_DISPLAY_NAMES } from '../../../shared/agent-launch/catalog/tui-agent-display-names' +import { AGENT_SESSION_ATTACHMENT_MAX_BYTES } from '../../../shared/agent-session/agent-session-attachments' import { claudeRecord } from './journal/claude-structured-item-translation' export class ClaudeDispatchContentError extends Error { @@ -51,10 +52,10 @@ export function claudeDispatchContentRejection(error: unknown): AgentJournalDisp return claudeDispatchRejection(agentSessionFailureFact('attachmentUnreadable')) } -const BYTES_PER_MB = 1024 * 1024 -const MAX_IMAGE_BYTES = 5 * BYTES_PER_MB +// Claude shrinks every image before the API sees it, so these bound host memory, not API limits. +const MAX_IMAGE_BYTES = AGENT_SESSION_ATTACHMENT_MAX_BYTES const MAX_IMAGE_COUNT = 20 -const MAX_TOTAL_IMAGE_BYTES = 20 * BYTES_PER_MB +const MAX_TOTAL_IMAGE_BYTES = 64 * 1024 * 1024 const MAX_REPLAY_CONTENT_KEY_BYTES = 256 type ImageBudget = { From 1cc94fdfd698c412af5d882985b292cf4a939412 Mon Sep 17 00:00:00 2001 From: OrcaWin Date: Wed, 11 Nov 2026 19:03:54 -0800 Subject: [PATCH 05/13] test: remove low-value workspace and launch suites (#41336) * test: remove low-value workspace and launch coverage * test: prune removed orphan cleanup timer baseline --------- Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com> --- config/flaky-test-pattern-baseline.txt | 2 +- config/reliability-gates.jsonc | 28 +- .../scripts/ci/sharding/ci-shard-timings.json | 4 - .../session-scanner-service-smoke.test.mjs | 49 ---- .../ipc/worktrees-delete-pty-teardown.test.ts | 192 ------------- .../worktrees-issue-command-overrides.test.ts | 217 --------------- ...worktrees-orphan-directory-cleanup.test.ts | 104 ------- ...orktrees-ssh-repo-owner-resolution.test.ts | 260 ------------------ .../at-rest-command-catalog-stamp.test.ts | 6 +- .../structured-chat-naming-title.test.ts | 22 -- .../opencode-root-model-overlay.test.ts | 18 -- 11 files changed, 16 insertions(+), 886 deletions(-) delete mode 100644 config/scripts/orcad-distribution/server-bundle/session-scanner-service-smoke.test.mjs delete mode 100644 src/main/ipc/worktrees-delete-pty-teardown.test.ts delete mode 100644 src/main/ipc/worktrees-issue-command-overrides.test.ts delete mode 100644 src/main/ipc/worktrees-ssh-repo-owner-resolution.test.ts diff --git a/config/flaky-test-pattern-baseline.txt b/config/flaky-test-pattern-baseline.txt index f785b8502aa..bdafee73c71 100644 --- a/config/flaky-test-pattern-baseline.txt +++ b/config/flaky-test-pattern-baseline.txt @@ -483,7 +483,7 @@ 2 test-no-wall-clock:src/main/grok/grok-hook-stdin-no-eof.test.ts 3 test-no-wall-clock:src/main/ipc/runtime-environments/runtime-environment-request-connections.test.ts 1 test-no-wall-clock:src/main/ipc/terminal-render-desync-evidence.test.ts -5 test-no-wall-clock:src/main/ipc/worktrees-orphan-directory-cleanup.test.ts +2 test-no-wall-clock:src/main/ipc/worktrees-orphan-directory-cleanup.test.ts 2 test-no-wall-clock:src/main/ipc/worktrees/git/worktree-git-status-attribution-real-git.test.ts 2 test-no-wall-clock:src/main/macos-permissions/macos-tcc-log-stream-orphans.test.ts 4 test-no-wall-clock:src/main/managed-server/activation/orcad-exited-own-fence-mutation-race.test.ts diff --git a/config/reliability-gates.jsonc b/config/reliability-gates.jsonc index 726d2c4ef77..a15b4a546f1 100644 --- a/config/reliability-gates.jsonc +++ b/config/reliability-gates.jsonc @@ -9284,7 +9284,7 @@ "providers": ["ssh", "remote-runtime", "wsl"], "coveredPlatforms": ["linux"], "coveredProviders": ["ssh", "remote-runtime"], - "coverageNotes": "Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3368 narrows stream-reconnect coverage to the actual ten synthetic transport epochs retaining one PTY identity, accepting and holding detached input for exactly-once later delivery, current decoded output and per-epoch subscription release. Prior rejection wording was inaccurate. Phase/backoff/cutoff/manual reconnect and parser/OSC/pause/retained-image/fatal/latch/claim/early-close alarms are intentionally lost without equivalent replacement. The Linux multi-pane reconnect receipt remains verbatim with its original scope; no fresh live SSH/provider/PTY/UI/mobile/mixed-version/cross-platform evidence is claimed. Audit3359 retires whole host-surface-replacement and create-handoff suites, removing their compound unit command tokens, test-file entries and assertion references. Fake polling/deadline/subscription, supplied stale-handle ownership, close and replacement examples are intentionally lost without equivalent replacement. The unrelated Linux paired multi-pane reconnect receipt stays verbatim with its original scope. No fresh SSH/provider/network/PTY recovery, historical-import, mobile, mixed-version or cross-platform validation is claimed. Audit3342 retires the complete client create-outcome identity/options/recovery suite and the malformed-or-contradictory repo-owner response matrix. Client replay identity, options, no-downgrade, quiescence/budget/manual retry, authoritative probe-error and malformed-owner refusal alarms are intentionally lost without equivalent replacement. Remaining repo-owner late catalog and old-host re-adoption boundaries do not replace them. The unrelated Linux paired-reconnect receipt stays verbatim with its original scope. Surviving selectors are unexecuted here; no fresh SSH/provider/process-deduplication, mobile, mixed-version or cross-platform validation is claimed. Audit3329 execution amendment: the reduced one-suite reconnect selector passed on Linux (1 case). Original source-review statements about no fresh execution describe the earlier static review; this new receipt validates only this one local deterministic unit case, not the complete gate or physical/runtime/rendered/provider/platform parity. Historical seven-case macOS and two-case Linux receipts remain archived verbatim. Audit3326 adopts reconciled source audit3324 and intentionally retires these low-value examples and their regression alarms: keeps a sidebar row for every still-live pane after a long sleep — Whole suite: 279 lines manually replay renderer ownership registry/status writes, snapshot patch/set, six-pane fixture inventory, and sidebar selectors/builders; expected pane IDs are production-computed. Does not run sleep/wake/transport or actual sidebar entry. Retires reconnect row-preservation assembly alarm.; still retires the regular TUI surface after an explicit terminal exit — Ordinary explicit-exit callback/map cleanup is repeated in the stale/new stream lifecycle case below. Retires separate regular-TUI baseline; keeper is not equivalent process/UI proof.; asks the HUB to recover an expired SSH pane and rebinds the host identity — Expired SSH owner error must ask the owning HUB to recover, then adopt fixed replacement handle, resubscribe and report rebind; actual transport dispatch and handle state prevent wrong-client recovery.; does not recover a pane whose relay reply was an identity mismatch — Identity mismatch means a different live PTY, so actual transport must surface refusal without recover/create, keeping original identity. Prevents destructive duplicate agent recovery.; fails closed when an older HUB cannot recover an expired SSH pane — Old HUB method_not_found is scripted mock refusal/error echo with no actual skewed host; repeats generic recovery failure propagation. Retires old-host recovery fallback refusal example.; names a held recovery's stopped WSL distro in readable copy, not the raw hold code — Pins exact WSL stopped-distro error prose after mock code; translation/copy assertion with no actual held recovery behavior. Retires readable-copy alarm.. Retained tests do not replace these alarms. Dated unrelated evidence remains scoped to its original execution; no historical source/import-closure equivalence, fresh execution, real network/provider/SSH, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Audit 3318 reduces runtime-environment-ssh-state.test.ts, intentionally losing metadata membership-pruning/new-target-only-read and other-environment/local-bucket isolation alarms. Preserve generation/failed-full-refresh assertion and unknown-target authoritative-list reread; these narrower cases are not equivalent replacements. The unchanged active two-case Linux reconnect materialization/sidebar receipt and covered platform retain their exact scope and do not validate the reduced SSH state selector. Preserve unrelated SSH assertions, selectors and archived history. No fresh SSH discovery, provider/PTY/network, rendered, mobile, mixed-version, cross-platform or full-gate execution is claimed. Audit3315 execution amendment: the reduced two-suite reconnect selector passed on Linux (2 cases). Original source-review statements about no fresh execution describe the earlier static review; this new receipt validates only these two local deterministic unit cases, not the complete gate or physical/runtime/rendered/provider/platform parity. Historical seven-case macOS receipt remains archived verbatim. Audit 3312 retires the ordinary six-pane distinct-live-handle reconnect/resubscription alarm from paired-reconnect-multi-pane-materialization.test.ts. The unchanged parked-pane activation/rematerialization case remains narrower and is not an equivalent replacement. The complete prior seven-contract macOS receipt is archived as historical evidence; it does not validate the reduced selector. No fresh socket, PTY, sleep/wake, provider, platform, mobile, mixed-version or rendered execution is claimed. Audit3096: remove the deleted reconnect-banner selector and its two UI aliases. The reduced recovery-state selector retains only the deadline cancelling pending backoff and explicit retry owning a new epoch alias; healthy-replacement stale-schedule and independent caller-cutoff cancellation aliases retire. Other retained state cases cover reentrant retry, owner settlement and same-epoch parked revival, without replacing those removed alarms. Prior admission-selector retirement, archived July 28 receipt, unrelated selectors/claims/receipts and old gaps remain unchanged; no fresh execution is claimed. Audit3086: retire the destroyed fresh-session fallback settlement/non-destructive reattach alias and its direct selector; retain the detach/attach selector and assertion unchanged. The retained admission cases cover adopt-before-data and explicitly buffered-death admission without a new spawn, without replacing fallback retirement. July 28 evidence was already archived by audit3077; no duplicate retirement or new execution is claimed. Audit2709 removes src/renderer/src/runtime/web-session-tabs-sync/inventory/remote-runtime-session-tabs-inflight.test.ts: intentionally retire waiting out an older inventory and sharing a fresh post-operation request among concurrent callers. No equivalent replacement or fresh execution is claimed. Dated receipts remain historical evidence for their unchanged selectors only; they do not validate the reduced commands or full current gate. Historical prior account (preserved verbatim): Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/ssh/connection/ssh-connection-manager.test.ts: tears a failed startup down quietly, so its error is not replaced by a disconnect; marks a dial that never reached the host, so setup reports SSH’s reason; removes the registration when bulk teardown fails; leaves a connection whose own reconnect is asking the user alone; clears the pool entry when the closed connection is still the registered one; still reuses a connection that is already up; dials a user's connection test. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2052 (sources2049/2050/2051): Remove complete low-value restored-pane-title and pane-visibility-identity suites and their current selectors. Their supplied snapshot/title and nullable identity-comparison alarms are intentionally lost without replacement. Unrelated dated evidenceRuns remain unchanged and do not validate the reduced commands. No fresh gate execution, actual PTY/provider, rendered, remote/mobile or cross-platform result is claimed. Historical prior account (preserved verbatim): Audit1983: Delete the terminal-create idempotency forwarding suite and its prospective selector. Preserve2 unrelated receipts without claiming the lost forwarding alarm was replaced. Changed-source prior results do not validate reduced-source bytes. No fresh execution, equivalent replacement, red/green, measured budget, soak, provider/PTY, rendered, live remote/mobile or cross-platform result is claimed. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Delete complete transport-unmount and queued-initial-CWD helper suites: despite lifecycle-shaped fixtures, the unmount owner returns a Boolean PTY-id predicate, while CWD cases check nullish field selection and callback counts. Lose their detach/destroy/teardown and one-shot-CWD supplied examples. Partial queued-pane-startup loses three identical/empty/nonowner identity examples, retaining four exact one-shot/replacement ownership callbacks. No assertion object maps to these three sources, so unrelated alias objects remain byte-exact. Preserve the current81ed prior removal of pane-split-startup; stale bd948 source metadata snapshots are not replayed. No actual process/PTY startup, transport shutdown, live SSH or rendered recovery is proven by this metadata change. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1925 (sources1922/1923): Delete the sole supplied public-SSH-state spread/sanitized-error projection and its provider-authority-pair alias. Also delete the complete tiny one-shot split-startup wrapper suite: callback-visible temporary field/reset, sequential payload and thrown-sentinel reset alarms disappear. No SSH provider epoch handoff, connection, process startup or actual terminal split was performed by either deleted suite. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1903 removes whole mocked detach, supplied geometry and/or close-request suites selected by this gate. Their pending-CWD/PTY-binding argument/order, fabricated hit-test and close bookkeeping alarms are intentionally lost without replacement. Valid remaining selectors/assertions and unchanged-selector execution receipts retain their original scope. Reduced unit selections have no fresh broad gate, real PTY/provider, rendered focus, or measured-budget validation. Historical prior account (preserved verbatim): Audit1645 removes low-value test projections, mock wiring, private-state assertions, inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained public behavior and refusal boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1640 removes low-value IPC, SSH, workspace and transcript-watcher tests without equivalent replacement coverage. Per-source reviews record retained public refusal, cancellation, execution-host and real-filesystem outcomes with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1342 changes referenced sources. Retained unchanged-source receipts apply only to their unchanged commands; the reduced selectors have no new execution result. Historical coverage account (not a current claim for changed sources): Audit1301 changes referenced sources. Retained unchanged-source receipts apply only to their unchanged commands; the reduced selectors have no new execution result. Historical coverage account (not a current claim for changed sources): Deterministic renderer coverage proves retained native and runtime SSH payloads are admitted through production routes only with valid complete authority, stale cleanup cannot unregister a replacement runtime terminal, direct SSH Git and folder panes keep their PTY bindings through disconnect and reconnect, one authority chain stops after two automatic attempts even when each timeout exceeds the rolling window, rejected acknowledgements mutate no store maps, and one shared exact attempt admits every concurrent split-pane spawn and reattach while preserving the first PTY as the tab fallback. A later sibling failure rotates the tab once, stale callbacks from the prior attempt mutate no state, split remount activity suppression is counted per leaf, primary PTY exit promotes a bound survivor or preserves an empty continuation gap for a late sibling, and primary, non-primary, or null-PTY detach preserves exact authority on both resulting tabs. Intentional pane disposal cancels its settlement timer without breaking StrictMode remount timeout ownership. A mounted remote-runtime terminal survives repeated transport partitions without changing PTY identity. A real encrypted-WebSocket oracle proves a successful reachability probe can replace a pre-ready shared-control socket without rejecting or duplicating the waiting RPC. Direct SSH coordinator tests cover immediate terminal finalization, hydration correction, damping, bounded retry, and telemetry non-interference. Client/server heartbeat tests cover timer suspension, socket generations fence stale callbacks, cold restored-terminal attachment retries, cached pixels remain unhealthy until authoritative replay, automatic retries stop after one minute, manual reconnect preserves the PTY, and pane closure releases recovery UI state. Current-transport recovery clears every error that transport surfaced without clearing or displaying a sibling pane's errors, while transport suppression, pane retention, and the rendered surface are independently bounded. Current macOS Electron journeys against an ephemeral Linux Docker SSH target cover exact-authority repo/worktree hydration, live terminal recovery after disconnect/reconnect, and eager six-terminal remount after renderer reload. A Windows remote-runtime smoke covers reachability and PTY round-trip. Multi-target live fanout, paired-close, WSL, and patched live partition journeys remain gaps. The test-value audit narrowed resolved-worktree-execution-host.test.ts to cache reuse and invalidation. Its direct hydrated-worktree, missing-folder, restored-owner, host-stamp and empty-runtime selection alarms are lost. Other routing and recovery suites do not establish equivalent selector coverage; historical run commands and results remain historical. The test-value audit removes standalone unwritten/released renderer-authority and cross-scope request-count examples. The retained synthetic mirror/store/sidebar and pending-inventory checks do not establish real PTY byte ownership, SSH reconnection or equivalent replacement coverage; historical commands and results remain historical. The test-value audit removes direct join/label ledgers, immediate failure and negative-only stale/removal/partial-authority examples, plus the standalone retained/live event projection. Retained synthetic store/hydration compositions do not establish real SSH continuity, remote authority delivery or equivalent replacement alarms. Historical commands and results remain historical. The renderer scheduling fixture retains in-flight coalescing only. Its immediate publication and negative stale-SSH-cleanup call ledgers were removed; the retained SSH recovery evidence does not replace those alarms. The accepted-handle-events observer suite is removed, including its scoped listener and settlement-release assertions. These observer alarms are lost; retained SSH/transport journeys are not equivalent replacement coverage. The test-value audit removed repeated remote-handle rotation inventories and snapshot-first pending-restart migration examples from terminal-pty-identity-replacement.test.ts, plus direct foreground metadata setter, confirmation, evidence-enum, and prefix/worktree cleanup examples. Those rotation, pending-restart, snapshot-first migration, and foreground scalar-update alarms are lost. The retained replacement case observes pane and tab identities in one subscriber publication; it does not establish equivalent migration, process-evidence, or input-routing coverage. Historical commands and results remain historical. The test-value audit narrows terminals-hydration.test.ts to a real reconnect batch with one measured store publication and hydration-seeded history consumed by a subsequent visit/Back check. Direct split-layout, runtime/folder/floating restoration, dead-leaf focus, cold/known-absent SSH deferral, title, default-tab marker, navigation initialization, and hydrationSucceeded scalar examples are removed; those specific alarms are lost. No actual target-removal transition, live SSH connection, session writer, or persisted restart is exercised by the retained cases. The old target-scoped/snapshot assertion references are retired because the current file does not contain those oracles. Other suites and historical runs do not establish equivalent replacement coverage. Audit1299 removes tests/e2e/terminal-cold-activation-deferral.spec.ts. Private-state/work-budget/assembled-fixture alarms are intentionally lost; retained suites are not equivalent replacement coverage. No new runtime/platform evidence is claimed. Audit1301: src/renderer/src/runtime/remote-runtime-session-tabs-inflight.test.ts reduced. Direct dedupe eviction/count and equal receipt-position sample loses authored alarm. No equivalent replacement coverage is claimed. Audit1318: src/shared/remote-runtime/control/remote-runtime-shared-control-connection.test.ts: Retain real local encrypted WebSocket regressions for pre-ready socket replacement preserving a request, close-before-ready logical subscription cleanup, unrelated RPC surviving another deadline, and pending setup abort leaving a sibling pending. Host server is synthetic and uses production crypto; kept scenarios also retain internal diagnostics/memory assertions unchanged. Remove happy routing, authority-envelope, unknown-ID/log, cleanup enum, binary, keepalive/liveness and production-budget/private socket inventories; intentionally lose those alarms, no physical host authorization, independent crypto or remote provider claim. Reduced selectors need fresh validation; historical full-suite receipts do not validate the changed coverage. No equivalent replacement or physical-provider claim. Audit1342: src/renderer/src/hooks/ipc-events/direct-ssh-connect-reply-routing.test.ts removed. Initial and event-triggered partial-authority hydration fixture alarms disappear. src/renderer/src/components/terminal-pane/terminal-error-accumulation.test.ts removed. Multiline dedup/storm bounds, resource classification/collision, scoped clear/reconnect strip and1000-pane map count alarms disappear. No equivalent replacement coverage is claimed. Audit1396: src/renderer/src/components/terminal-pane/pty-connection-pending-spawn.spec.ts reduced. Lose standalone debug-state retention/log silence alarm. React StrictMode is not mounted; bindings manually simulate remount over mocked transport. No equivalent replacement coverage is claimed. The existing primary-selector receipt and its claims remain scoped to their unchanged sources; the broad optional command has no new execution result. Audit1400: src/main/ipc/ssh/ssh.test.ts reduced. Lose trivial missing target and mocked publication inventory. Preexisting unresolved ../ssh/ssh-port-scanner mock is recorded, not fixed. src/renderer/src/runtime/sync-runtime-graph-scheduling.test.ts removed. Removes whole coalescing call-count suite; no newest-state delivery or replacement claim. src/renderer/src/store/slices/terminal-pty-identity-replacement.test.ts reduced. Removes three no-op/work-count registrations; retains externally observed ownership consistency. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result. Audit1412: src/main/ssh/ssh-provider-authority.test.ts removed. Loses real AbortController revocation/target isolation/reentrancy outcomes, recreated-target generation-floor/session rollover behavior, stale mutation refusal and settled-registration cleanup alarms. These useful alarms are deliberately lost; no equivalent cancellation coverage is claimed. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result. Audit1743 removes the standalone line-height clamp/default and pane cwd fallback helper tests (5 authored cases). These trivial-helper alarms are intentionally lost, with no equivalent replacement. Only current commands/testFiles lose the deleted selectors; historical receipts and evidenceRuns remain unchanged and do not validate the reduced commands.", + "coverageNotes": "Audit3554 intentionally removes the whole worktrees-ssh-repo-owner-resolution suite (2 authored cases). Dedicated duplicate-exact/ambiguous-legacy owner refusal and fixed runtime-host/wrong-target/missing-authority/zero-owner/contradictory-owner status alarms are lost. The suite did not exercise an authority transition or actual SSH listing; its no-local-Git spy was also unreachable on an admitted SSH metadata-only route. Remaining suites are not equivalent replacements. Current selectors have no fresh execution from this static packet; historical receipts retain their original scope. No new SSH/provider/native/UI/mobile/mixed-version/cross-platform result is claimed. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3368 narrows stream-reconnect coverage to the actual ten synthetic transport epochs retaining one PTY identity, accepting and holding detached input for exactly-once later delivery, current decoded output and per-epoch subscription release. Prior rejection wording was inaccurate. Phase/backoff/cutoff/manual reconnect and parser/OSC/pause/retained-image/fatal/latch/claim/early-close alarms are intentionally lost without equivalent replacement. The Linux multi-pane reconnect receipt remains verbatim with its original scope; no fresh live SSH/provider/PTY/UI/mobile/mixed-version/cross-platform evidence is claimed. Audit3359 retires whole host-surface-replacement and create-handoff suites, removing their compound unit command tokens, test-file entries and assertion references. Fake polling/deadline/subscription, supplied stale-handle ownership, close and replacement examples are intentionally lost without equivalent replacement. The unrelated Linux paired multi-pane reconnect receipt stays verbatim with its original scope. No fresh SSH/provider/network/PTY recovery, historical-import, mobile, mixed-version or cross-platform validation is claimed. Audit3342 retires the complete client create-outcome identity/options/recovery suite and the malformed-or-contradictory repo-owner response matrix. Client replay identity, options, no-downgrade, quiescence/budget/manual retry, authoritative probe-error and malformed-owner refusal alarms are intentionally lost without equivalent replacement. Remaining repo-owner late catalog and old-host re-adoption boundaries do not replace them. The unrelated Linux paired-reconnect receipt stays verbatim with its original scope. Surviving selectors are unexecuted here; no fresh SSH/provider/process-deduplication, mobile, mixed-version or cross-platform validation is claimed. Audit3329 execution amendment: the reduced one-suite reconnect selector passed on Linux (1 case). Original source-review statements about no fresh execution describe the earlier static review; this new receipt validates only this one local deterministic unit case, not the complete gate or physical/runtime/rendered/provider/platform parity. Historical seven-case macOS and two-case Linux receipts remain archived verbatim. Audit3326 adopts reconciled source audit3324 and intentionally retires these low-value examples and their regression alarms: keeps a sidebar row for every still-live pane after a long sleep — Whole suite: 279 lines manually replay renderer ownership registry/status writes, snapshot patch/set, six-pane fixture inventory, and sidebar selectors/builders; expected pane IDs are production-computed. Does not run sleep/wake/transport or actual sidebar entry. Retires reconnect row-preservation assembly alarm.; still retires the regular TUI surface after an explicit terminal exit — Ordinary explicit-exit callback/map cleanup is repeated in the stale/new stream lifecycle case below. Retires separate regular-TUI baseline; keeper is not equivalent process/UI proof.; asks the HUB to recover an expired SSH pane and rebinds the host identity — Expired SSH owner error must ask the owning HUB to recover, then adopt fixed replacement handle, resubscribe and report rebind; actual transport dispatch and handle state prevent wrong-client recovery.; does not recover a pane whose relay reply was an identity mismatch — Identity mismatch means a different live PTY, so actual transport must surface refusal without recover/create, keeping original identity. Prevents destructive duplicate agent recovery.; fails closed when an older HUB cannot recover an expired SSH pane — Old HUB method_not_found is scripted mock refusal/error echo with no actual skewed host; repeats generic recovery failure propagation. Retires old-host recovery fallback refusal example.; names a held recovery's stopped WSL distro in readable copy, not the raw hold code — Pins exact WSL stopped-distro error prose after mock code; translation/copy assertion with no actual held recovery behavior. Retires readable-copy alarm.. Retained tests do not replace these alarms. Dated unrelated evidence remains scoped to its original execution; no historical source/import-closure equivalence, fresh execution, real network/provider/SSH, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Audit 3318 reduces runtime-environment-ssh-state.test.ts, intentionally losing metadata membership-pruning/new-target-only-read and other-environment/local-bucket isolation alarms. Preserve generation/failed-full-refresh assertion and unknown-target authoritative-list reread; these narrower cases are not equivalent replacements. The unchanged active two-case Linux reconnect materialization/sidebar receipt and covered platform retain their exact scope and do not validate the reduced SSH state selector. Preserve unrelated SSH assertions, selectors and archived history. No fresh SSH discovery, provider/PTY/network, rendered, mobile, mixed-version, cross-platform or full-gate execution is claimed. Audit3315 execution amendment: the reduced two-suite reconnect selector passed on Linux (2 cases). Original source-review statements about no fresh execution describe the earlier static review; this new receipt validates only these two local deterministic unit cases, not the complete gate or physical/runtime/rendered/provider/platform parity. Historical seven-case macOS receipt remains archived verbatim. Audit 3312 retires the ordinary six-pane distinct-live-handle reconnect/resubscription alarm from paired-reconnect-multi-pane-materialization.test.ts. The unchanged parked-pane activation/rematerialization case remains narrower and is not an equivalent replacement. The complete prior seven-contract macOS receipt is archived as historical evidence; it does not validate the reduced selector. No fresh socket, PTY, sleep/wake, provider, platform, mobile, mixed-version or rendered execution is claimed. Audit3096: remove the deleted reconnect-banner selector and its two UI aliases. The reduced recovery-state selector retains only the deadline cancelling pending backoff and explicit retry owning a new epoch alias; healthy-replacement stale-schedule and independent caller-cutoff cancellation aliases retire. Other retained state cases cover reentrant retry, owner settlement and same-epoch parked revival, without replacing those removed alarms. Prior admission-selector retirement, archived July 28 receipt, unrelated selectors/claims/receipts and old gaps remain unchanged; no fresh execution is claimed. Audit3086: retire the destroyed fresh-session fallback settlement/non-destructive reattach alias and its direct selector; retain the detach/attach selector and assertion unchanged. The retained admission cases cover adopt-before-data and explicitly buffered-death admission without a new spawn, without replacing fallback retirement. July 28 evidence was already archived by audit3077; no duplicate retirement or new execution is claimed. Audit2709 removes src/renderer/src/runtime/web-session-tabs-sync/inventory/remote-runtime-session-tabs-inflight.test.ts: intentionally retire waiting out an older inventory and sharing a fresh post-operation request among concurrent callers. No equivalent replacement or fresh execution is claimed. Dated receipts remain historical evidence for their unchanged selectors only; they do not validate the reduced commands or full current gate. Historical prior account (preserved verbatim): Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/ssh/connection/ssh-connection-manager.test.ts: tears a failed startup down quietly, so its error is not replaced by a disconnect; marks a dial that never reached the host, so setup reports SSH’s reason; removes the registration when bulk teardown fails; leaves a connection whose own reconnect is asking the user alone; clears the pool entry when the closed connection is still the registered one; still reuses a connection that is already up; dials a user's connection test. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2052 (sources2049/2050/2051): Remove complete low-value restored-pane-title and pane-visibility-identity suites and their current selectors. Their supplied snapshot/title and nullable identity-comparison alarms are intentionally lost without replacement. Unrelated dated evidenceRuns remain unchanged and do not validate the reduced commands. No fresh gate execution, actual PTY/provider, rendered, remote/mobile or cross-platform result is claimed. Historical prior account (preserved verbatim): Audit1983: Delete the terminal-create idempotency forwarding suite and its prospective selector. Preserve2 unrelated receipts without claiming the lost forwarding alarm was replaced. Changed-source prior results do not validate reduced-source bytes. No fresh execution, equivalent replacement, red/green, measured budget, soak, provider/PTY, rendered, live remote/mobile or cross-platform result is claimed. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Delete complete transport-unmount and queued-initial-CWD helper suites: despite lifecycle-shaped fixtures, the unmount owner returns a Boolean PTY-id predicate, while CWD cases check nullish field selection and callback counts. Lose their detach/destroy/teardown and one-shot-CWD supplied examples. Partial queued-pane-startup loses three identical/empty/nonowner identity examples, retaining four exact one-shot/replacement ownership callbacks. No assertion object maps to these three sources, so unrelated alias objects remain byte-exact. Preserve the current81ed prior removal of pane-split-startup; stale bd948 source metadata snapshots are not replayed. No actual process/PTY startup, transport shutdown, live SSH or rendered recovery is proven by this metadata change. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1925 (sources1922/1923): Delete the sole supplied public-SSH-state spread/sanitized-error projection and its provider-authority-pair alias. Also delete the complete tiny one-shot split-startup wrapper suite: callback-visible temporary field/reset, sequential payload and thrown-sentinel reset alarms disappear. No SSH provider epoch handoff, connection, process startup or actual terminal split was performed by either deleted suite. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1903 removes whole mocked detach, supplied geometry and/or close-request suites selected by this gate. Their pending-CWD/PTY-binding argument/order, fabricated hit-test and close bookkeeping alarms are intentionally lost without replacement. Valid remaining selectors/assertions and unchanged-selector execution receipts retain their original scope. Reduced unit selections have no fresh broad gate, real PTY/provider, rendered focus, or measured-budget validation. Historical prior account (preserved verbatim): Audit1645 removes low-value test projections, mock wiring, private-state assertions, inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained public behavior and refusal boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1640 removes low-value IPC, SSH, workspace and transcript-watcher tests without equivalent replacement coverage. Per-source reviews record retained public refusal, cancellation, execution-host and real-filesystem outcomes with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1342 changes referenced sources. Retained unchanged-source receipts apply only to their unchanged commands; the reduced selectors have no new execution result. Historical coverage account (not a current claim for changed sources): Audit1301 changes referenced sources. Retained unchanged-source receipts apply only to their unchanged commands; the reduced selectors have no new execution result. Historical coverage account (not a current claim for changed sources): Deterministic renderer coverage proves retained native and runtime SSH payloads are admitted through production routes only with valid complete authority, stale cleanup cannot unregister a replacement runtime terminal, direct SSH Git and folder panes keep their PTY bindings through disconnect and reconnect, one authority chain stops after two automatic attempts even when each timeout exceeds the rolling window, rejected acknowledgements mutate no store maps, and one shared exact attempt admits every concurrent split-pane spawn and reattach while preserving the first PTY as the tab fallback. A later sibling failure rotates the tab once, stale callbacks from the prior attempt mutate no state, split remount activity suppression is counted per leaf, primary PTY exit promotes a bound survivor or preserves an empty continuation gap for a late sibling, and primary, non-primary, or null-PTY detach preserves exact authority on both resulting tabs. Intentional pane disposal cancels its settlement timer without breaking StrictMode remount timeout ownership. A mounted remote-runtime terminal survives repeated transport partitions without changing PTY identity. A real encrypted-WebSocket oracle proves a successful reachability probe can replace a pre-ready shared-control socket without rejecting or duplicating the waiting RPC. Direct SSH coordinator tests cover immediate terminal finalization, hydration correction, damping, bounded retry, and telemetry non-interference. Client/server heartbeat tests cover timer suspension, socket generations fence stale callbacks, cold restored-terminal attachment retries, cached pixels remain unhealthy until authoritative replay, automatic retries stop after one minute, manual reconnect preserves the PTY, and pane closure releases recovery UI state. Current-transport recovery clears every error that transport surfaced without clearing or displaying a sibling pane's errors, while transport suppression, pane retention, and the rendered surface are independently bounded. Current macOS Electron journeys against an ephemeral Linux Docker SSH target cover exact-authority repo/worktree hydration, live terminal recovery after disconnect/reconnect, and eager six-terminal remount after renderer reload. A Windows remote-runtime smoke covers reachability and PTY round-trip. Multi-target live fanout, paired-close, WSL, and patched live partition journeys remain gaps. The test-value audit narrowed resolved-worktree-execution-host.test.ts to cache reuse and invalidation. Its direct hydrated-worktree, missing-folder, restored-owner, host-stamp and empty-runtime selection alarms are lost. Other routing and recovery suites do not establish equivalent selector coverage; historical run commands and results remain historical. The test-value audit removes standalone unwritten/released renderer-authority and cross-scope request-count examples. The retained synthetic mirror/store/sidebar and pending-inventory checks do not establish real PTY byte ownership, SSH reconnection or equivalent replacement coverage; historical commands and results remain historical. The test-value audit removes direct join/label ledgers, immediate failure and negative-only stale/removal/partial-authority examples, plus the standalone retained/live event projection. Retained synthetic store/hydration compositions do not establish real SSH continuity, remote authority delivery or equivalent replacement alarms. Historical commands and results remain historical. The renderer scheduling fixture retains in-flight coalescing only. Its immediate publication and negative stale-SSH-cleanup call ledgers were removed; the retained SSH recovery evidence does not replace those alarms. The accepted-handle-events observer suite is removed, including its scoped listener and settlement-release assertions. These observer alarms are lost; retained SSH/transport journeys are not equivalent replacement coverage. The test-value audit removed repeated remote-handle rotation inventories and snapshot-first pending-restart migration examples from terminal-pty-identity-replacement.test.ts, plus direct foreground metadata setter, confirmation, evidence-enum, and prefix/worktree cleanup examples. Those rotation, pending-restart, snapshot-first migration, and foreground scalar-update alarms are lost. The retained replacement case observes pane and tab identities in one subscriber publication; it does not establish equivalent migration, process-evidence, or input-routing coverage. Historical commands and results remain historical. The test-value audit narrows terminals-hydration.test.ts to a real reconnect batch with one measured store publication and hydration-seeded history consumed by a subsequent visit/Back check. Direct split-layout, runtime/folder/floating restoration, dead-leaf focus, cold/known-absent SSH deferral, title, default-tab marker, navigation initialization, and hydrationSucceeded scalar examples are removed; those specific alarms are lost. No actual target-removal transition, live SSH connection, session writer, or persisted restart is exercised by the retained cases. The old target-scoped/snapshot assertion references are retired because the current file does not contain those oracles. Other suites and historical runs do not establish equivalent replacement coverage. Audit1299 removes tests/e2e/terminal-cold-activation-deferral.spec.ts. Private-state/work-budget/assembled-fixture alarms are intentionally lost; retained suites are not equivalent replacement coverage. No new runtime/platform evidence is claimed. Audit1301: src/renderer/src/runtime/remote-runtime-session-tabs-inflight.test.ts reduced. Direct dedupe eviction/count and equal receipt-position sample loses authored alarm. No equivalent replacement coverage is claimed. Audit1318: src/shared/remote-runtime/control/remote-runtime-shared-control-connection.test.ts: Retain real local encrypted WebSocket regressions for pre-ready socket replacement preserving a request, close-before-ready logical subscription cleanup, unrelated RPC surviving another deadline, and pending setup abort leaving a sibling pending. Host server is synthetic and uses production crypto; kept scenarios also retain internal diagnostics/memory assertions unchanged. Remove happy routing, authority-envelope, unknown-ID/log, cleanup enum, binary, keepalive/liveness and production-budget/private socket inventories; intentionally lose those alarms, no physical host authorization, independent crypto or remote provider claim. Reduced selectors need fresh validation; historical full-suite receipts do not validate the changed coverage. No equivalent replacement or physical-provider claim. Audit1342: src/renderer/src/hooks/ipc-events/direct-ssh-connect-reply-routing.test.ts removed. Initial and event-triggered partial-authority hydration fixture alarms disappear. src/renderer/src/components/terminal-pane/terminal-error-accumulation.test.ts removed. Multiline dedup/storm bounds, resource classification/collision, scoped clear/reconnect strip and1000-pane map count alarms disappear. No equivalent replacement coverage is claimed. Audit1396: src/renderer/src/components/terminal-pane/pty-connection-pending-spawn.spec.ts reduced. Lose standalone debug-state retention/log silence alarm. React StrictMode is not mounted; bindings manually simulate remount over mocked transport. No equivalent replacement coverage is claimed. The existing primary-selector receipt and its claims remain scoped to their unchanged sources; the broad optional command has no new execution result. Audit1400: src/main/ipc/ssh/ssh.test.ts reduced. Lose trivial missing target and mocked publication inventory. Preexisting unresolved ../ssh/ssh-port-scanner mock is recorded, not fixed. src/renderer/src/runtime/sync-runtime-graph-scheduling.test.ts removed. Removes whole coalescing call-count suite; no newest-state delivery or replacement claim. src/renderer/src/store/slices/terminal-pty-identity-replacement.test.ts reduced. Removes three no-op/work-count registrations; retains externally observed ownership consistency. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result. Audit1412: src/main/ssh/ssh-provider-authority.test.ts removed. Loses real AbortController revocation/target isolation/reentrancy outcomes, recreated-target generation-floor/session rollover behavior, stale mutation refusal and settled-registration cleanup alarms. These useful alarms are deliberately lost; no equivalent cancellation coverage is claimed. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result. Audit1743 removes the standalone line-height clamp/default and pane cwd fallback helper tests (5 authored cases). These trivial-helper alarms are intentionally lost, with no equivalent replacement. Only current commands/testFiles lose the deleted selectors; historical receipts and evidenceRuns remain unchanged and do not validate the reduced commands.", "motivatingLinks": [ "https://linear.app/stably/issue/STA-3107", "https://github.com/stablyai/orca/pull/12664", @@ -9306,7 +9306,7 @@ "pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/components/terminal-pane/pty-transport/ipc/pty-transport-detach-attach-handoff.test.ts", "pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-stream-reconnect.test.ts src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-expired-pane-recovery.test.ts src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/recovery/remote-runtime-pty-recovery-state.test.ts src/shared/remote-runtime/remote-runtime-socket-liveness.test.ts src/shared/remote-runtime/control/remote-runtime-shared-control-connection.test.ts src/shared/remote-runtime/control/remote-runtime-shared-control-socket-generation.test.ts src/shared/remote-runtime/remote-runtime-client-error-classification.test.ts src/main/runtime/rpc/remote-runtime-server-heartbeat.test.ts src/main/runtime/terminal/launch/orca-runtime-terminal-create-idempotency.test.ts", "pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/store/slices/ui/workspace-navigation-history-hydration.test.ts src/renderer/src/store/repos/repos-ssh-host-reconciliation.test.ts --reporter=dot", - "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/ssh/ssh.test.ts src/main/ipc/worktrees-ssh-repo-owner-resolution.test.ts src/main/ssh/connection/ssh-connection-manager.test.ts src/main/ssh/connection/ssh-connection.test.ts src/preload/ssh-authority-forwarding.test.ts src/renderer/src/runtime/rpc/subscriptions/runtime-client-events.test.ts src/renderer/src/runtime/hosts/ssh/runtime-environment-ssh-state.test.ts src/shared/ssh-retained-payload-admission.test.ts --reporter=dot", + "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/ssh/ssh.test.ts src/main/ssh/connection/ssh-connection-manager.test.ts src/main/ssh/connection/ssh-connection.test.ts src/preload/ssh-authority-forwarding.test.ts src/renderer/src/runtime/rpc/subscriptions/runtime-client-events.test.ts src/renderer/src/runtime/hosts/ssh/runtime-environment-ssh-state.test.ts src/shared/ssh-retained-payload-admission.test.ts --reporter=dot", "pnpm exec electron-vite build --mode e2e", "pnpm run build:web-from-renderer", "SKIP_BUILD=1 pnpm exec playwright test tests/e2e/paired-remote-terminal-materialization-reconnect.spec.ts --config tests/playwright.config.ts --project electron-headless --workers=1" @@ -9331,7 +9331,6 @@ "src/renderer/src/store/slices/ui/workspace-navigation-history-hydration.test.ts", "src/renderer/src/store/repos/repos-ssh-host-reconciliation.test.ts", "src/main/ipc/ssh/ssh.test.ts", - "src/main/ipc/worktrees-ssh-repo-owner-resolution.test.ts", "src/main/ssh/connection/ssh-connection-manager.test.ts", "src/main/ssh/connection/ssh-connection.test.ts", "src/preload/ssh-authority-forwarding.test.ts", @@ -9534,7 +9533,8 @@ "Audit3359 historical original assertionRefs object, preserved verbatim; retired claims are not active evidence or replacement coverage:\n{\n \"file\": \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-create-handoff.test.ts\",\n \"assertions\": [\n \"a stale create completion cannot replace or close a newer attachment, including when two runtimes use the same raw handle\"\n ]\n }", "Audit3368 narrows stream-reconnect coverage to the actual ten synthetic transport epochs retaining one PTY identity, accepting and holding detached input for exactly-once later delivery, current decoded output and per-epoch subscription release. Prior rejection wording was inaccurate. Phase/backoff/cutoff/manual reconnect and parser/OSC/pause/retained-image/fatal/latch/claim/early-close alarms are intentionally lost without equivalent replacement. The Linux multi-pane reconnect receipt remains verbatim with its original scope; no fresh live SSH/provider/PTY/UI/mobile/mixed-version/cross-platform evidence is claimed.", "Audit3368 historical original assertionRefs object, preserved verbatim; retired claims are not active evidence or replacement coverage:\n{\n \"file\": \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-stream-reconnect.test.ts\",\n \"assertions\": [\n \"recovery exposes connecting, recovering, backoff, connected, and fatal-offline phases with fenced epochs\",\n \"ten partition cycles retain one PTY identity, reject detached input, and unsubscribe each epoch exactly once\",\n \"automatic retries stop after one minute with no fatal error, no further requests, and no stale input\",\n \"manual reconnect starts a new epoch, resubscribes the same PTY exactly once, and becomes healthy only after its snapshot\"\n ]\n }", - "Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed.\nOriginal active selector fields (historical raw native JSON values; not current coverage):\ncommands: [\n \"pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/paired-reconnect-multi-pane-materialization.test.ts --reporter=dot\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/components/terminal/background-terminal-worktree-mount.test.ts src/renderer/src/components/terminal-pane/mounting/queued-pane-startup.test.ts src/renderer/src/components/terminal-pane/parking/one-shot-park-replay.test.ts src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-spawn-and-startup.test.ts src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-stream-reconnect.test.ts src/renderer/src/store/terminals/terminal-pty-identity-replacement.test.ts\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/components/terminal-pane/pty-transport/ipc/pty-transport-detach-attach-handoff.test.ts\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-stream-reconnect.test.ts src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-expired-pane-recovery.test.ts src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/recovery/remote-runtime-pty-recovery-state.test.ts src/shared/remote-runtime/remote-runtime-socket-liveness.test.ts src/shared/remote-runtime/control/remote-runtime-shared-control-connection.test.ts src/shared/remote-runtime/control/remote-runtime-shared-control-socket-generation.test.ts src/shared/remote-runtime/remote-runtime-client-error-classification.test.ts src/main/runtime/rpc/remote-runtime-server-heartbeat.test.ts src/main/runtime/terminal/launch/orca-runtime-terminal-create-idempotency.test.ts\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/store/slices/ui/workspace-navigation-history-hydration.test.ts src/renderer/src/store/repos/repos-ssh-host-reconciliation.test.ts src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-spawn-and-startup.test.ts --reporter=dot\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/ssh/ssh.test.ts src/main/ipc/worktrees-ssh-repo-owner-resolution.test.ts src/main/ssh/connection/ssh-connection-manager.test.ts src/main/ssh/connection/ssh-connection.test.ts src/preload/ssh-authority-forwarding.test.ts src/renderer/src/runtime/rpc/subscriptions/runtime-client-events.test.ts src/renderer/src/runtime/hosts/ssh/runtime-environment-ssh-state.test.ts src/shared/ssh-retained-payload-admission.test.ts --reporter=dot\",\n \"pnpm exec electron-vite build --mode e2e\",\n \"pnpm run build:web-from-renderer\",\n \"SKIP_BUILD=1 pnpm exec playwright test tests/e2e/paired-remote-terminal-materialization-reconnect.spec.ts --config tests/playwright.config.ts --project electron-headless --workers=1\"\n ]\ntestFiles: [\n \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/paired-reconnect-multi-pane-materialization.test.ts\",\n \"src/renderer/src/components/terminal/background-terminal-worktree-mount.test.ts\",\n \"src/renderer/src/components/terminal-pane/mounting/queued-pane-startup.test.ts\",\n \"src/renderer/src/components/terminal-pane/parking/one-shot-park-replay.test.ts\",\n \"src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-spawn-and-startup.test.ts\",\n \"src/renderer/src/components/terminal-pane/pty-transport/ipc/pty-transport-detach-attach-handoff.test.ts\",\n \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-stream-reconnect.test.ts\",\n \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-expired-pane-recovery.test.ts\",\n \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/recovery/remote-runtime-pty-recovery-state.test.ts\",\n \"src/renderer/src/store/terminals/terminal-pty-identity-replacement.test.ts\",\n \"src/shared/remote-runtime/remote-runtime-socket-liveness.test.ts\",\n \"src/shared/remote-runtime/control/remote-runtime-shared-control-connection.test.ts\",\n \"src/shared/remote-runtime/control/remote-runtime-shared-control-socket-generation.test.ts\",\n \"src/shared/remote-runtime/remote-runtime-client-error-classification.test.ts\",\n \"src/main/runtime/rpc/remote-runtime-server-heartbeat.test.ts\",\n \"src/main/runtime/terminal/launch/orca-runtime-terminal-create-idempotency.test.ts\",\n \"tests/e2e/paired-remote-terminal-materialization-reconnect.spec.ts\",\n \"src/renderer/src/store/slices/ui/workspace-navigation-history-hydration.test.ts\",\n \"src/renderer/src/store/repos/repos-ssh-host-reconciliation.test.ts\",\n \"src/main/ipc/ssh/ssh.test.ts\",\n \"src/main/ipc/worktrees-ssh-repo-owner-resolution.test.ts\",\n \"src/main/ssh/connection/ssh-connection-manager.test.ts\",\n \"src/main/ssh/connection/ssh-connection.test.ts\",\n \"src/preload/ssh-authority-forwarding.test.ts\",\n \"src/renderer/src/runtime/rpc/subscriptions/runtime-client-events.test.ts\",\n \"src/renderer/src/runtime/hosts/ssh/runtime-environment-ssh-state.test.ts\",\n \"src/shared/ssh-retained-payload-admission.test.ts\"\n ]\nassertionRefs: [\n {\n \"file\": \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/paired-reconnect-multi-pane-materialization.test.ts\",\n \"assertions\": [\n \"a pane parked while the client slept is authoritatively activated and rematerialized without leaving any pane handle null\"\n ]\n },\n {\n \"file\": \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-stream-reconnect.test.ts\",\n \"assertions\": [\n \"ten synthetic partition cycles retain one PTY identity, deliver each held detached input exactly once and release each epoch subscription exactly once; decoded live input/output remains current\"\n ]\n },\n {\n \"file\": \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/recovery/remote-runtime-pty-recovery-state.test.ts\",\n \"assertions\": [\n \"the automatic-recovery deadline cancels pending backoff and a manual retry owns a new epoch\"\n ]\n },\n {\n \"file\": \"src/shared/remote-runtime/remote-runtime-socket-liveness.test.ts\",\n \"assertions\": [\n \"a suspended client receives a fresh probe deadline after resume and post-resume activity clears it\"\n ]\n },\n {\n \"file\": \"src/main/runtime/terminal/launch/orca-runtime-terminal-create-idempotency.test.ts\",\n \"assertions\": [\n \"one authenticated client/worktree mutation derives a stable server-owned terminal handle\",\n \"an unknown-outcome retry adopts the same provider PTY after runtime-process restart without rerunning startup\",\n \"retry inventory failure, missing same-worktree identity metadata, or ownership mismatch fails closed without spawning\",\n \"in-flight mutation coordination is bounded and releases capacity after settlement\"\n ]\n },\n {\n \"file\": \"src/main/ssh/connection/ssh-connection-manager.test.ts\",\n \"assertions\": [\n \"disconnect invalidates a pending transport attempt immediately so late rejection or resolution cannot remove the replacement\"\n ]\n },\n {\n \"file\": \"src/main/ssh/connection/ssh-connection.test.ts\",\n \"assertions\": [\n \"late ssh2 ready and startup error events after disconnect cannot resurrect or overwrite disconnected state\"\n ]\n },\n {\n \"file\": \"src/preload/ssh-authority-forwarding.test.ts\",\n \"assertions\": [\n \"full authority crosses Electron IPC without loss\",\n \"partial authority becomes unknown for bounded reconciliation and malformed full authority is dropped\",\n \"variable-form host-qualified worktree requests retain fail-closed outcomes in their return type\"\n ]\n },\n {\n \"file\": \"src/shared/ssh-retained-payload-admission.test.ts\",\n \"assertions\": [\n \"retained connection states reject partial or malformed authority\",\n \"only partial compatibility authority can normalize to unknown for bounded direct-SSH reconciliation\",\n \"shared direct SSH authority admission requires bounded identifiers and a non-negative safe generation\"\n ]\n },\n {\n \"file\": \"src/renderer/src/runtime/rpc/subscriptions/runtime-client-events.test.ts\",\n \"assertions\": [\n \"partial runtime authority is rejected through the captured subscription callback before environment state receives it\"\n ]\n },\n {\n \"file\": \"src/renderer/src/runtime/hosts/ssh/runtime-environment-ssh-state.test.ts\",\n \"assertions\": [\n \"held replies from an old generation cannot replace a newer generation; failed full refresh preserves live state behind queued metadata\",\n \"an unknown target triggers an authoritative list reread\"\n ]\n },\n {\n \"file\": \"src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-spawn-and-startup.test.ts\",\n \"assertions\": [\n \"a StrictMode remount joins the pending spawn instead of starting a second one\"\n ]\n },\n {\n \"file\": \"src/renderer/src/components/terminal-pane/pty-transport/ipc/pty-transport-detach-attach-handoff.test.ts\",\n \"assertions\": [\n \"abandoning an obsolete reattach drops its data, replay, write-unavailable, and exit handlers without killing the durable PTY\"\n ]\n }\n ]" + "Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed.\nOriginal active selector fields (historical raw native JSON values; not current coverage):\ncommands: [\n \"pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/paired-reconnect-multi-pane-materialization.test.ts --reporter=dot\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/components/terminal/background-terminal-worktree-mount.test.ts src/renderer/src/components/terminal-pane/mounting/queued-pane-startup.test.ts src/renderer/src/components/terminal-pane/parking/one-shot-park-replay.test.ts src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-spawn-and-startup.test.ts src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-stream-reconnect.test.ts src/renderer/src/store/terminals/terminal-pty-identity-replacement.test.ts\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/components/terminal-pane/pty-transport/ipc/pty-transport-detach-attach-handoff.test.ts\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-stream-reconnect.test.ts src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-expired-pane-recovery.test.ts src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/recovery/remote-runtime-pty-recovery-state.test.ts src/shared/remote-runtime/remote-runtime-socket-liveness.test.ts src/shared/remote-runtime/control/remote-runtime-shared-control-connection.test.ts src/shared/remote-runtime/control/remote-runtime-shared-control-socket-generation.test.ts src/shared/remote-runtime/remote-runtime-client-error-classification.test.ts src/main/runtime/rpc/remote-runtime-server-heartbeat.test.ts src/main/runtime/terminal/launch/orca-runtime-terminal-create-idempotency.test.ts\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/store/slices/ui/workspace-navigation-history-hydration.test.ts src/renderer/src/store/repos/repos-ssh-host-reconciliation.test.ts src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-spawn-and-startup.test.ts --reporter=dot\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/ssh/ssh.test.ts src/main/ipc/worktrees-ssh-repo-owner-resolution.test.ts src/main/ssh/connection/ssh-connection-manager.test.ts src/main/ssh/connection/ssh-connection.test.ts src/preload/ssh-authority-forwarding.test.ts src/renderer/src/runtime/rpc/subscriptions/runtime-client-events.test.ts src/renderer/src/runtime/hosts/ssh/runtime-environment-ssh-state.test.ts src/shared/ssh-retained-payload-admission.test.ts --reporter=dot\",\n \"pnpm exec electron-vite build --mode e2e\",\n \"pnpm run build:web-from-renderer\",\n \"SKIP_BUILD=1 pnpm exec playwright test tests/e2e/paired-remote-terminal-materialization-reconnect.spec.ts --config tests/playwright.config.ts --project electron-headless --workers=1\"\n ]\ntestFiles: [\n \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/paired-reconnect-multi-pane-materialization.test.ts\",\n \"src/renderer/src/components/terminal/background-terminal-worktree-mount.test.ts\",\n \"src/renderer/src/components/terminal-pane/mounting/queued-pane-startup.test.ts\",\n \"src/renderer/src/components/terminal-pane/parking/one-shot-park-replay.test.ts\",\n \"src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-spawn-and-startup.test.ts\",\n \"src/renderer/src/components/terminal-pane/pty-transport/ipc/pty-transport-detach-attach-handoff.test.ts\",\n \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-stream-reconnect.test.ts\",\n \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-expired-pane-recovery.test.ts\",\n \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/recovery/remote-runtime-pty-recovery-state.test.ts\",\n \"src/renderer/src/store/terminals/terminal-pty-identity-replacement.test.ts\",\n \"src/shared/remote-runtime/remote-runtime-socket-liveness.test.ts\",\n \"src/shared/remote-runtime/control/remote-runtime-shared-control-connection.test.ts\",\n \"src/shared/remote-runtime/control/remote-runtime-shared-control-socket-generation.test.ts\",\n \"src/shared/remote-runtime/remote-runtime-client-error-classification.test.ts\",\n \"src/main/runtime/rpc/remote-runtime-server-heartbeat.test.ts\",\n \"src/main/runtime/terminal/launch/orca-runtime-terminal-create-idempotency.test.ts\",\n \"tests/e2e/paired-remote-terminal-materialization-reconnect.spec.ts\",\n \"src/renderer/src/store/slices/ui/workspace-navigation-history-hydration.test.ts\",\n \"src/renderer/src/store/repos/repos-ssh-host-reconciliation.test.ts\",\n \"src/main/ipc/ssh/ssh.test.ts\",\n \"src/main/ipc/worktrees-ssh-repo-owner-resolution.test.ts\",\n \"src/main/ssh/connection/ssh-connection-manager.test.ts\",\n \"src/main/ssh/connection/ssh-connection.test.ts\",\n \"src/preload/ssh-authority-forwarding.test.ts\",\n \"src/renderer/src/runtime/rpc/subscriptions/runtime-client-events.test.ts\",\n \"src/renderer/src/runtime/hosts/ssh/runtime-environment-ssh-state.test.ts\",\n \"src/shared/ssh-retained-payload-admission.test.ts\"\n ]\nassertionRefs: [\n {\n \"file\": \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/paired-reconnect-multi-pane-materialization.test.ts\",\n \"assertions\": [\n \"a pane parked while the client slept is authoritatively activated and rematerialized without leaving any pane handle null\"\n ]\n },\n {\n \"file\": \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-stream-reconnect.test.ts\",\n \"assertions\": [\n \"ten synthetic partition cycles retain one PTY identity, deliver each held detached input exactly once and release each epoch subscription exactly once; decoded live input/output remains current\"\n ]\n },\n {\n \"file\": \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/recovery/remote-runtime-pty-recovery-state.test.ts\",\n \"assertions\": [\n \"the automatic-recovery deadline cancels pending backoff and a manual retry owns a new epoch\"\n ]\n },\n {\n \"file\": \"src/shared/remote-runtime/remote-runtime-socket-liveness.test.ts\",\n \"assertions\": [\n \"a suspended client receives a fresh probe deadline after resume and post-resume activity clears it\"\n ]\n },\n {\n \"file\": \"src/main/runtime/terminal/launch/orca-runtime-terminal-create-idempotency.test.ts\",\n \"assertions\": [\n \"one authenticated client/worktree mutation derives a stable server-owned terminal handle\",\n \"an unknown-outcome retry adopts the same provider PTY after runtime-process restart without rerunning startup\",\n \"retry inventory failure, missing same-worktree identity metadata, or ownership mismatch fails closed without spawning\",\n \"in-flight mutation coordination is bounded and releases capacity after settlement\"\n ]\n },\n {\n \"file\": \"src/main/ssh/connection/ssh-connection-manager.test.ts\",\n \"assertions\": [\n \"disconnect invalidates a pending transport attempt immediately so late rejection or resolution cannot remove the replacement\"\n ]\n },\n {\n \"file\": \"src/main/ssh/connection/ssh-connection.test.ts\",\n \"assertions\": [\n \"late ssh2 ready and startup error events after disconnect cannot resurrect or overwrite disconnected state\"\n ]\n },\n {\n \"file\": \"src/preload/ssh-authority-forwarding.test.ts\",\n \"assertions\": [\n \"full authority crosses Electron IPC without loss\",\n \"partial authority becomes unknown for bounded reconciliation and malformed full authority is dropped\",\n \"variable-form host-qualified worktree requests retain fail-closed outcomes in their return type\"\n ]\n },\n {\n \"file\": \"src/shared/ssh-retained-payload-admission.test.ts\",\n \"assertions\": [\n \"retained connection states reject partial or malformed authority\",\n \"only partial compatibility authority can normalize to unknown for bounded direct-SSH reconciliation\",\n \"shared direct SSH authority admission requires bounded identifiers and a non-negative safe generation\"\n ]\n },\n {\n \"file\": \"src/renderer/src/runtime/rpc/subscriptions/runtime-client-events.test.ts\",\n \"assertions\": [\n \"partial runtime authority is rejected through the captured subscription callback before environment state receives it\"\n ]\n },\n {\n \"file\": \"src/renderer/src/runtime/hosts/ssh/runtime-environment-ssh-state.test.ts\",\n \"assertions\": [\n \"held replies from an old generation cannot replace a newer generation; failed full refresh preserves live state behind queued metadata\",\n \"an unknown target triggers an authoritative list reread\"\n ]\n },\n {\n \"file\": \"src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-spawn-and-startup.test.ts\",\n \"assertions\": [\n \"a StrictMode remount joins the pending spawn instead of starting a second one\"\n ]\n },\n {\n \"file\": \"src/renderer/src/components/terminal-pane/pty-transport/ipc/pty-transport-detach-attach-handoff.test.ts\",\n \"assertions\": [\n \"abandoning an obsolete reattach drops its data, replay, write-unavailable, and exit handlers without killing the durable PTY\"\n ]\n }\n ]", + "Audit3554 intentionally removes the whole worktrees-ssh-repo-owner-resolution suite (2 authored cases). Dedicated duplicate-exact/ambiguous-legacy owner refusal and fixed runtime-host/wrong-target/missing-authority/zero-owner/contradictory-owner status alarms are lost. The suite did not exercise an authority transition or actual SSH listing; its no-local-Git spy was also unreachable on an admitted SSH metadata-only route. Remaining suites are not equivalent replacements. Current selectors have no fresh execution from this static packet; historical receipts retain their original scope. No new SSH/provider/native/UI/mobile/mixed-version/cross-platform result is claimed." ], "demotionRule": "Cannot promote if provider failure can close panes or if the oracle is screenshot-only." }, @@ -12329,22 +12329,22 @@ "providers": ["local", "remote-runtime", "ssh"], "coveredPlatforms": ["macos", "linux"], "coveredProviders": ["local", "remote-runtime", "ssh"], - "coverageNotes": "Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3460 final-v2 composes accepted AUTHOR3454/PEER3455 on actual post263 main: intentionally removes99 authored/106 potential expanded cases across16 selected drivers, including7 whole suites, retaining24 authored/26 potential cases. No equivalent replacement or fresh runtime is claimed. Broad root-unit discovery loses whole cross-client workspace-sync, terminal-preview, picker-skill-hint and close-routing suites plus reduced worker-stop, remote-owned-record, scanner, mobile-bundle and daemon-health samples. Separate E2E discovery also loses the paired mobile/desktop mirror journey and two unit drivers. Retained callback/table/outer-loop/guard bytes are exact; retained checks do not replace the removed alarms. No app, native device, mixed-version, remote or cross-platform execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The retained runtime registration driver loses three complete imported fragments: changed-working-label title facts and suppressed-frame tracked snapshots; refusing SSH GitLab MR-base lookup before local Git/glab calls; and missing folder-workspace path creation/update refusal before spawn/store mutation. These alarms are intentionally lost. Other imported fragments and separately rooted dated receipts are unchanged; they are not equivalent replacement coverage. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2090 (sources2088/2089): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2086 (sources2082/2083): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-handles-part-02.spec.ts: recovers a moved SSH pane through the tab it now sits in, not the one its lease froze; reports a plain expired orphan; does not report a pane whose only recent lease was superseded; does not report a pane whose only recent lease had its relay id recycled; picks the eligible orphan over a superseded predecessor that matches first; collects the same tabs the per-tab reader reports, in one sweep of the leases; reports no candidate when no lease names any of the worktree tabs. src/main/runtime/orca-runtime-tests/runtime-availability.spec.ts: starts unavailable with no authoritative window; reports the configured Windows terminal shell on status; reports floating workspace availability from settings on status; polls floating tabs with targeted PTY liveness and no repo/provider inventory; hydrates persisted tabs when the store cannot report repos; advertises browser screencast only when a renderer window is available; routes mobile Codex reset consumption through the account mutation coordinator; maps a definite pre-provider rejection into an authoritative current snapshot. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2056 (sources2054/2055): Intentionally retire these low-value alarms without replacement: src/renderer/src/runtime/files/watching/runtime-file-client-watch.test.ts: delivers remote file changes to listeners for the same worktree. src/main/runtime/orca-runtime.test.ts: ordinary registration cleanup removes only await import('./orca-runtime-tests/mobile-session-tabs-part-14.spec'). src/main/runtime/orca-runtime-tests/terminal-side-effect-facts-part-03.spec.ts: arms the stale-title timer for a seeded working title; arms the Command Code scrape from the noted spawn command; prefers the tracked title over a stale renderer lastTitle in the hydration seed; writes recent output into the current runtime output map; uses a later title handler replacement with the runtime receiver; calls the current injected listener with the runtime receiver; keeps optional reconciliation lazy after its callback is removed. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-14.spec.ts: lists them, so a paired client can mirror a migrated editor; closes one by retiring it from the host session; keeps an unsaved draft unless the close is forced; shares aggregate work but leaves a replacement pending promise alone; lets a targeted refresh finish while an aggregate refresh waits; refuses a close when the store lacks %s before the guard; keeps a later missing %s on the original caught failure path. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Exact retired assertionRefs, preserved as historical objects only: {\n \"file\": \"src/renderer/src/runtime/files/watching/runtime-file-client-watch.test.ts\",\n \"assertions\": [\n \"terminal watcher error and setup failure evict the shared entry before retrying callbacks run\",\n \"an old stream end cannot evict the replacement stream, which still delivers to its listener\",\n \"two same-worktree listeners share a synthetic subscription until the final stop unsubscribes and sends unwatch\"\n ]\n } Historical prior account (preserved verbatim): Audit1973: Delete the standalone mocked watcher cleanup suite and its entire sole-file prospective command. Retire the removed file-command transport-abort alias; preserve the independent cancellation suite and final-overflow/error/end cleanup callback, and13 unrelated receipt objects. Whole original changed dated receipts are retired, never narrowed or transferred. This metadata review claims no fresh execution, equivalent replacement, provider/PTY, live remote/mobile, rendered, cross-platform, soak or measured-budget result. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Partial runtime files-watch suite retains controlled Windows close-event waiting, shared-client survival, late physical-exit ownership/deadline cleanup and shutdown joining. Its combined Windows close/deadline/late-exit alias remains backed by unchanged callbacks. Retire native-error-close, pre-publication setup ownership and failed-delete rearm aliases; associated delegation, failure forwarding, removal routing and recovery examples are lost. These tests inject watch/subscription seams and do not observe an actual native watcher or physical process. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1925 (sources1922/1923): Partial runtime watcher-host suite loses mocked option/event/overflow/resubscription forwarding, supplied teardown-error propagation, aborted-signal/no-call and private ten-thousand-waiter-count inventories. Retire the 10,000-caller bound and failed-recovery-subscriber-ending assertion aliases. Keep same-root owner cancellation, pre-ready replacement, held root-close join, stale-generation recovery delivery and initial failed-root ownership mappings backed by retained real-host callbacks with injected subscription/error/physical-exit promises. No actual native watcher, process termination or resource bound was observed. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1683 removes low-value test projections, mock wiring, private-state assertions, inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained public behavior and refusal boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1645 removes low-value test projections, mock wiring, private-state assertions, inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained public behavior and refusal boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1640 removes low-value IPC, SSH, workspace and transcript-watcher tests without equivalent replacement coverage. Per-source reviews record retained public refusal, cancellation, execution-host and real-filesystem outcomes with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1634 removes low-value IPC, filesystem and runtime projections, argument inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained cancellation, ownership, authorization, filesystem and SQLite boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1627 removes low-value mocked daemon projections, scalar tables and duplicate terminal examples without equivalent replacement coverage. Per-source reviews record retained terminal content preservation, process ownership refusal and recovery boundaries together with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1623 removes low-value mocked daemon projections, scalar tables and duplicate terminal examples without equivalent replacement coverage. Per-source reviews record retained terminal content preservation, process ownership refusal and recovery boundaries together with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1516 removes package-electron-runtime-contract.test.mjs, a copied release-workflow platform and marker matrix. Its workflow wiring alarm is intentionally retired; no actual packaged watcher delivery was exercised and no equivalent replacement is claimed. Unaffected dated execution receipts retain their original scope; changed optional selectors have no fresh full-gate execution result. This audit removes the paired-web stopped-during-outage and late-ready direct-branch examples. The retained pre-ready cancellation, failed-unwatch retry and replay-continuity cases are not equivalent replacements for those ordering alarms. Current web-runtime-client.test.ts assertions below name only its retained pre-ready callback case; replay cases remain in their separate suite. Historical commands and results remain historical. Deterministic tests cover one shared healthy child per runtime process plus at most four bounded fault-quarantine children, native desktop, paired-runtime, and WSL snapshot processes sharing the same eight-physical-child reservation with typed serialized event-driven capacity recovery, including recursive re-wait when crash recovery reclaims an announced slot, one quarantine attempt per watch lifetime, generation-scoped cancellation, bounded termination with removable deadline waiters, replacement-crawl errors that cannot be blessed by a late readiness ack, pre-ready paired-web cancellation registered before an unbounded capacity wait, already-resolved and late physical-exit retry of a rejected paired-web teardown, physical desktop-install and local/SSH PTY teardown, stale-generation rejection, per-child stat bounds, final-overflow RPC delivery, and renderer eviction before terminal callbacks can retry. Destructive local and SSH removal fences both ID-derived and resolved-cwd terminal roots, including sibling-root/cwd combinations with reverse admission rollback, closes descendant relay watches before parent deletion, enumerates authoritative provider/cwd ownership, falls back to daemon spawn cwd before OSC 7, and keeps daemon/relay immediate shutdown pending until native exit or a bounded fail-closed error; a dead relay PID discovered during attach also settles concurrent shutdown before its stale entry is reaped, while Windows runtime watcher deadlines retain one removable physical-close waiter, treat Node's error-close path as positive physical-exit proof, and clear root ownership on late close. The relay policy test maps a standard repository's base, Git common directory, and worktree roots to one healthy supervisor, then proves a shared-child failure recovers those roots in separate quarantine supervisors with overflow and resumed events. Abortable capacity, quarantine, runtime-root, relay, SSH, and relay pre-install setup waits attach one reaction to each shared promise and explicitly remove 10,000 cancelled caller closures while one anchor remains. Built-entry macOS harnesses kill the desktop/runtime and SSH relay watcher children, then require automatic resubscription and later events while the host process and relay PTY survive. SSH bundles the child boundary beside relay.js, requires both artifacts for install completeness, and preserves registration-owned same-root cancellation. The test-value audit corrects the active renderer watch command/file reference to the existing runtime-file-client-watch.test.ts, whose three cases are retained unchanged. They observe synthetic subscription retry/fanout and unsubscribe/unwatch calls, not actual preload/main bookkeeping or physical watcher-process cleanup. The separate read/preview client suite loses forwarding, supplied-error and negative-only examples, retaining binary detection and Unicode SVG decoding. Historical evidence commands/results are unchanged. Test-value audit1293 removes src/main/providers/local-pty/local-pty-provider-session-inventory.test.ts (2 authored cases). Their isolated assertions no longer run; surviving tests are not claimed as equivalent coverage. Fresh execution against the candidate SHA is required. Audit1312 removes low-value imported specs from aggregate test drivers; old whole-driver execution receipts are archived. Retained registrations are unchanged. No equivalent replacement coverage is claimed. Audit1383: src/main/runtime/orca-runtime-tests/worktree-native-chat-setup.spec.ts reduced. Lose command-list/receipt/default/background activation matrices whose provider spawn returns canned IDs. Retained test proves actual tab allocation separation, not terminal startup. No equivalent replacement coverage is claimed. Existing exact unaffected receipts and their platform/provider/budget claims remain scoped to unchanged selectors. Changed optional selectors have no new execution result. Audit1401: src/main/runtime/orca-runtime-tests/mobile-session-tabs-duplicate-pty-repair.spec.ts removed. Loses listing-on-already-repaired-data alarm; it never established runtime repair invocation. src/main/runtime/orca-runtime-tests/mobile-session-tabs-split-agent-tab.spec.ts removed. Loses actual mobile split sibling shell versus recorded-agent resume command alarms; no equivalent coverage asserted. src/main/runtime/orca-runtime-tests/worktree-native-chat-setup.spec.ts removed. Loses separate setup/startup tab result in that manually assembled path. src/main/runtime/orca-runtime.test.ts reduced. Driver registration/import of a whole-deleted selected spec is removed; directly authored cases remain unchanged. No extra authored removal credit. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result. Audit1412: src/main/workspaces/removal/watcher-removal-gate.test.ts reduced. Loses exact removal-error compatibility with renderer predicate under shared message constants. Kept public refusal outcomes do not establish that renderer classification. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result.", + "coverageNotes": "Audit3553/3559 current scope: delete worktrees-delete-pty-teardown.test.ts and worktrees-issue-command-overrides.test.ts entirely and remove three cases from worktrees-orphan-directory-cleanup.test.ts (6 authored/6 finite removals). Intentionally retire typed unserved-SSH create/delete admission, exact-name collision/error-token, proven-orphan no-force confirmation, standalone .git-directory refusal, and SSH-owned local-path preservation alarms. Retain only actual authorized orphan deletion and no-marker notes-file preservation; they are not equivalent replacements. The no-marker fixture resolves the default getRepos owner, so it does not prove temporary owner-repo binding. Git/Store/PTY/notifications are mocked; no physical process stop, durable metadata, real SSH delivery or native UI proof is claimed. Unchanged historical oracle, coverage, evidenceRuns and escaped receipts retain their original scopes only; no fresh execution or transferred pass validates the reduced sources. Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3460 final-v2 composes accepted AUTHOR3454/PEER3455 on actual post263 main: intentionally removes99 authored/106 potential expanded cases across16 selected drivers, including7 whole suites, retaining24 authored/26 potential cases. No equivalent replacement or fresh runtime is claimed. Broad root-unit discovery loses whole cross-client workspace-sync, terminal-preview, picker-skill-hint and close-routing suites plus reduced worker-stop, remote-owned-record, scanner, mobile-bundle and daemon-health samples. Separate E2E discovery also loses the paired mobile/desktop mirror journey and two unit drivers. Retained callback/table/outer-loop/guard bytes are exact; retained checks do not replace the removed alarms. No app, native device, mixed-version, remote or cross-platform execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The retained runtime registration driver loses three complete imported fragments: changed-working-label title facts and suppressed-frame tracked snapshots; refusing SSH GitLab MR-base lookup before local Git/glab calls; and missing folder-workspace path creation/update refusal before spawn/store mutation. These alarms are intentionally lost. Other imported fragments and separately rooted dated receipts are unchanged; they are not equivalent replacement coverage. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2090 (sources2088/2089): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2086 (sources2082/2083): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-handles-part-02.spec.ts: recovers a moved SSH pane through the tab it now sits in, not the one its lease froze; reports a plain expired orphan; does not report a pane whose only recent lease was superseded; does not report a pane whose only recent lease had its relay id recycled; picks the eligible orphan over a superseded predecessor that matches first; collects the same tabs the per-tab reader reports, in one sweep of the leases; reports no candidate when no lease names any of the worktree tabs. src/main/runtime/orca-runtime-tests/runtime-availability.spec.ts: starts unavailable with no authoritative window; reports the configured Windows terminal shell on status; reports floating workspace availability from settings on status; polls floating tabs with targeted PTY liveness and no repo/provider inventory; hydrates persisted tabs when the store cannot report repos; advertises browser screencast only when a renderer window is available; routes mobile Codex reset consumption through the account mutation coordinator; maps a definite pre-provider rejection into an authoritative current snapshot. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2056 (sources2054/2055): Intentionally retire these low-value alarms without replacement: src/renderer/src/runtime/files/watching/runtime-file-client-watch.test.ts: delivers remote file changes to listeners for the same worktree. src/main/runtime/orca-runtime.test.ts: ordinary registration cleanup removes only await import('./orca-runtime-tests/mobile-session-tabs-part-14.spec'). src/main/runtime/orca-runtime-tests/terminal-side-effect-facts-part-03.spec.ts: arms the stale-title timer for a seeded working title; arms the Command Code scrape from the noted spawn command; prefers the tracked title over a stale renderer lastTitle in the hydration seed; writes recent output into the current runtime output map; uses a later title handler replacement with the runtime receiver; calls the current injected listener with the runtime receiver; keeps optional reconciliation lazy after its callback is removed. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-14.spec.ts: lists them, so a paired client can mirror a migrated editor; closes one by retiring it from the host session; keeps an unsaved draft unless the close is forced; shares aggregate work but leaves a replacement pending promise alone; lets a targeted refresh finish while an aggregate refresh waits; refuses a close when the store lacks %s before the guard; keeps a later missing %s on the original caught failure path. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Exact retired assertionRefs, preserved as historical objects only: {\n \"file\": \"src/renderer/src/runtime/files/watching/runtime-file-client-watch.test.ts\",\n \"assertions\": [\n \"terminal watcher error and setup failure evict the shared entry before retrying callbacks run\",\n \"an old stream end cannot evict the replacement stream, which still delivers to its listener\",\n \"two same-worktree listeners share a synthetic subscription until the final stop unsubscribes and sends unwatch\"\n ]\n } Historical prior account (preserved verbatim): Audit1973: Delete the standalone mocked watcher cleanup suite and its entire sole-file prospective command. Retire the removed file-command transport-abort alias; preserve the independent cancellation suite and final-overflow/error/end cleanup callback, and13 unrelated receipt objects. Whole original changed dated receipts are retired, never narrowed or transferred. This metadata review claims no fresh execution, equivalent replacement, provider/PTY, live remote/mobile, rendered, cross-platform, soak or measured-budget result. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Partial runtime files-watch suite retains controlled Windows close-event waiting, shared-client survival, late physical-exit ownership/deadline cleanup and shutdown joining. Its combined Windows close/deadline/late-exit alias remains backed by unchanged callbacks. Retire native-error-close, pre-publication setup ownership and failed-delete rearm aliases; associated delegation, failure forwarding, removal routing and recovery examples are lost. These tests inject watch/subscription seams and do not observe an actual native watcher or physical process. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1925 (sources1922/1923): Partial runtime watcher-host suite loses mocked option/event/overflow/resubscription forwarding, supplied teardown-error propagation, aborted-signal/no-call and private ten-thousand-waiter-count inventories. Retire the 10,000-caller bound and failed-recovery-subscriber-ending assertion aliases. Keep same-root owner cancellation, pre-ready replacement, held root-close join, stale-generation recovery delivery and initial failed-root ownership mappings backed by retained real-host callbacks with injected subscription/error/physical-exit promises. No actual native watcher, process termination or resource bound was observed. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1683 removes low-value test projections, mock wiring, private-state assertions, inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained public behavior and refusal boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1645 removes low-value test projections, mock wiring, private-state assertions, inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained public behavior and refusal boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1640 removes low-value IPC, SSH, workspace and transcript-watcher tests without equivalent replacement coverage. Per-source reviews record retained public refusal, cancellation, execution-host and real-filesystem outcomes with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1634 removes low-value IPC, filesystem and runtime projections, argument inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained cancellation, ownership, authorization, filesystem and SQLite boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1627 removes low-value mocked daemon projections, scalar tables and duplicate terminal examples without equivalent replacement coverage. Per-source reviews record retained terminal content preservation, process ownership refusal and recovery boundaries together with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1623 removes low-value mocked daemon projections, scalar tables and duplicate terminal examples without equivalent replacement coverage. Per-source reviews record retained terminal content preservation, process ownership refusal and recovery boundaries together with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1516 removes package-electron-runtime-contract.test.mjs, a copied release-workflow platform and marker matrix. Its workflow wiring alarm is intentionally retired; no actual packaged watcher delivery was exercised and no equivalent replacement is claimed. Unaffected dated execution receipts retain their original scope; changed optional selectors have no fresh full-gate execution result. This audit removes the paired-web stopped-during-outage and late-ready direct-branch examples. The retained pre-ready cancellation, failed-unwatch retry and replay-continuity cases are not equivalent replacements for those ordering alarms. Current web-runtime-client.test.ts assertions below name only its retained pre-ready callback case; replay cases remain in their separate suite. Historical commands and results remain historical. Deterministic tests cover one shared healthy child per runtime process plus at most four bounded fault-quarantine children, native desktop, paired-runtime, and WSL snapshot processes sharing the same eight-physical-child reservation with typed serialized event-driven capacity recovery, including recursive re-wait when crash recovery reclaims an announced slot, one quarantine attempt per watch lifetime, generation-scoped cancellation, bounded termination with removable deadline waiters, replacement-crawl errors that cannot be blessed by a late readiness ack, pre-ready paired-web cancellation registered before an unbounded capacity wait, already-resolved and late physical-exit retry of a rejected paired-web teardown, physical desktop-install and local/SSH PTY teardown, stale-generation rejection, per-child stat bounds, final-overflow RPC delivery, and renderer eviction before terminal callbacks can retry. Destructive local and SSH removal fences both ID-derived and resolved-cwd terminal roots, including sibling-root/cwd combinations with reverse admission rollback, closes descendant relay watches before parent deletion, enumerates authoritative provider/cwd ownership, falls back to daemon spawn cwd before OSC 7, and keeps daemon/relay immediate shutdown pending until native exit or a bounded fail-closed error; a dead relay PID discovered during attach also settles concurrent shutdown before its stale entry is reaped, while Windows runtime watcher deadlines retain one removable physical-close waiter, treat Node's error-close path as positive physical-exit proof, and clear root ownership on late close. The relay policy test maps a standard repository's base, Git common directory, and worktree roots to one healthy supervisor, then proves a shared-child failure recovers those roots in separate quarantine supervisors with overflow and resumed events. Abortable capacity, quarantine, runtime-root, relay, SSH, and relay pre-install setup waits attach one reaction to each shared promise and explicitly remove 10,000 cancelled caller closures while one anchor remains. Built-entry macOS harnesses kill the desktop/runtime and SSH relay watcher children, then require automatic resubscription and later events while the host process and relay PTY survive. SSH bundles the child boundary beside relay.js, requires both artifacts for install completeness, and preserves registration-owned same-root cancellation. The test-value audit corrects the active renderer watch command/file reference to the existing runtime-file-client-watch.test.ts, whose three cases are retained unchanged. They observe synthetic subscription retry/fanout and unsubscribe/unwatch calls, not actual preload/main bookkeeping or physical watcher-process cleanup. The separate read/preview client suite loses forwarding, supplied-error and negative-only examples, retaining binary detection and Unicode SVG decoding. Historical evidence commands/results are unchanged. Test-value audit1293 removes src/main/providers/local-pty/local-pty-provider-session-inventory.test.ts (2 authored cases). Their isolated assertions no longer run; surviving tests are not claimed as equivalent coverage. Fresh execution against the candidate SHA is required. Audit1312 removes low-value imported specs from aggregate test drivers; old whole-driver execution receipts are archived. Retained registrations are unchanged. No equivalent replacement coverage is claimed. Audit1383: src/main/runtime/orca-runtime-tests/worktree-native-chat-setup.spec.ts reduced. Lose command-list/receipt/default/background activation matrices whose provider spawn returns canned IDs. Retained test proves actual tab allocation separation, not terminal startup. No equivalent replacement coverage is claimed. Existing exact unaffected receipts and their platform/provider/budget claims remain scoped to unchanged selectors. Changed optional selectors have no new execution result. Audit1401: src/main/runtime/orca-runtime-tests/mobile-session-tabs-duplicate-pty-repair.spec.ts removed. Loses listing-on-already-repaired-data alarm; it never established runtime repair invocation. src/main/runtime/orca-runtime-tests/mobile-session-tabs-split-agent-tab.spec.ts removed. Loses actual mobile split sibling shell versus recorded-agent resume command alarms; no equivalent coverage asserted. src/main/runtime/orca-runtime-tests/worktree-native-chat-setup.spec.ts removed. Loses separate setup/startup tab result in that manually assembled path. src/main/runtime/orca-runtime.test.ts reduced. Driver registration/import of a whole-deleted selected spec is removed; directly authored cases remain unchanged. No extra authored removal credit. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result. Audit1412: src/main/workspaces/removal/watcher-removal-gate.test.ts reduced. Loses exact removal-error compatibility with renderer predicate under shared message constants. Kept public refusal outcomes do not establish that renderer classification. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result.", "motivatingLinks": [ "https://github.com/stablyai/orca/issues/5308", "https://github.com/stablyai/orca/issues/7547", "https://github.com/stablyai/orca/issues/8212" ], "invariant": "Local runtime and SSH relay roots share one watcher child per host process while healthy and use at most four quarantine children after shard failure; native desktop, paired-runtime, and direct WSL snapshot children consume the same process-wide eight-child reservation and retry transient exhaustion without permanently disabling a root. Cancellation remains owned by its child generation and settles only after acknowledgement or bounded physical exit; paired-web retains cleanup ownership and can cancel before readiness even while capacity is exhausted, retries a failed unwatch after already-resolved or late physical exit without duplicating stream termination, replacements never overlap a still-retiring generation, and aborted or timed-out callers detach immediately even while queued behind same-root relay teardown. Destructive local and SSH cleanup holds one containment-aware watcher/terminal removal fence through descendant watcher and PTY physical exit plus parent Git/filesystem deletion; daemon and relay kill requests retain ownership until the actual subprocess exit callback or an equivalent positive dead-PID proof and propagate rejected signals or a bounded failure otherwise. A crash-fused shared shard moves affected roots into the bounded quarantine pool, a second fused quarantine ends those watch lifetimes, interrupted setup emits overflow after recovery, metadata work stays bounded, and the SSH relay, PTYs, Git, and other filesystem operations never load the watcher native addon or share its fault boundary.", - "oracle": "Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3460 final-v2 composes accepted AUTHOR3454/PEER3455 on actual post263 main: intentionally removes99 authored/106 potential expanded cases across16 selected drivers, including7 whole suites, retaining24 authored/26 potential cases. No equivalent replacement or fresh runtime is claimed. Broad root-unit discovery loses whole cross-client workspace-sync, terminal-preview, picker-skill-hint and close-routing suites plus reduced worker-stop, remote-owned-record, scanner, mobile-bundle and daemon-health samples. Separate E2E discovery also loses the paired mobile/desktop mirror journey and two unit drivers. Retained callback/table/outer-loop/guard bytes are exact; retained checks do not replace the removed alarms. No app, native device, mixed-version, remote or cross-platform execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The retained runtime registration driver loses three complete imported fragments: changed-working-label title facts and suppressed-frame tracked snapshots; refusing SSH GitLab MR-base lookup before local Git/glab calls; and missing folder-workspace path creation/update refusal before spawn/store mutation. These alarms are intentionally lost. Other imported fragments and separately rooted dated receipts are unchanged; they are not equivalent replacement coverage. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2090 (sources2088/2089): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2086 (sources2082/2083): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-handles-part-02.spec.ts: recovers a moved SSH pane through the tab it now sits in, not the one its lease froze; reports a plain expired orphan; does not report a pane whose only recent lease was superseded; does not report a pane whose only recent lease had its relay id recycled; picks the eligible orphan over a superseded predecessor that matches first; collects the same tabs the per-tab reader reports, in one sweep of the leases; reports no candidate when no lease names any of the worktree tabs. src/main/runtime/orca-runtime-tests/runtime-availability.spec.ts: starts unavailable with no authoritative window; reports the configured Windows terminal shell on status; reports floating workspace availability from settings on status; polls floating tabs with targeted PTY liveness and no repo/provider inventory; hydrates persisted tabs when the store cannot report repos; advertises browser screencast only when a renderer window is available; routes mobile Codex reset consumption through the account mutation coordinator; maps a definite pre-provider rejection into an authoritative current snapshot. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2056 (sources2054/2055): Intentionally retire these low-value alarms without replacement: src/renderer/src/runtime/files/watching/runtime-file-client-watch.test.ts: delivers remote file changes to listeners for the same worktree. src/main/runtime/orca-runtime.test.ts: ordinary registration cleanup removes only await import('./orca-runtime-tests/mobile-session-tabs-part-14.spec'). src/main/runtime/orca-runtime-tests/terminal-side-effect-facts-part-03.spec.ts: arms the stale-title timer for a seeded working title; arms the Command Code scrape from the noted spawn command; prefers the tracked title over a stale renderer lastTitle in the hydration seed; writes recent output into the current runtime output map; uses a later title handler replacement with the runtime receiver; calls the current injected listener with the runtime receiver; keeps optional reconciliation lazy after its callback is removed. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-14.spec.ts: lists them, so a paired client can mirror a migrated editor; closes one by retiring it from the host session; keeps an unsaved draft unless the close is forced; shares aggregate work but leaves a replacement pending promise alone; lets a targeted refresh finish while an aggregate refresh waits; refuses a close when the store lacks %s before the guard; keeps a later missing %s on the original caught failure path. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Exact retired assertionRefs, preserved as historical objects only: {\n \"file\": \"src/renderer/src/runtime/files/watching/runtime-file-client-watch.test.ts\",\n \"assertions\": [\n \"terminal watcher error and setup failure evict the shared entry before retrying callbacks run\",\n \"an old stream end cannot evict the replacement stream, which still delivers to its listener\",\n \"two same-worktree listeners share a synthetic subscription until the final stop unsubscribes and sends unwatch\"\n ]\n } Historical prior account (preserved verbatim): Audit1973: Delete the standalone mocked watcher cleanup suite and its entire sole-file prospective command. Retire the removed file-command transport-abort alias; preserve the independent cancellation suite and final-overflow/error/end cleanup callback, and13 unrelated receipt objects. Whole original changed dated receipts are retired, never narrowed or transferred. This metadata review claims no fresh execution, equivalent replacement, provider/PTY, live remote/mobile, rendered, cross-platform, soak or measured-budget result. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Partial runtime files-watch suite retains controlled Windows close-event waiting, shared-client survival, late physical-exit ownership/deadline cleanup and shutdown joining. Its combined Windows close/deadline/late-exit alias remains backed by unchanged callbacks. Retire native-error-close, pre-publication setup ownership and failed-delete rearm aliases; associated delegation, failure forwarding, removal routing and recovery examples are lost. These tests inject watch/subscription seams and do not observe an actual native watcher or physical process. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1925 (sources1922/1923): Partial runtime watcher-host suite loses mocked option/event/overflow/resubscription forwarding, supplied teardown-error propagation, aborted-signal/no-call and private ten-thousand-waiter-count inventories. Retire the 10,000-caller bound and failed-recovery-subscriber-ending assertion aliases. Keep same-root owner cancellation, pre-ready replacement, held root-close join, stale-generation recovery delivery and initial failed-root ownership mappings backed by retained real-host callbacks with injected subscription/error/physical-exit promises. No actual native watcher, process termination or resource bound was observed. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1683 retires changed-source assertion and execution scope without equivalent replacement coverage. Existing unchanged-selector receipts remain scoped to their original sources. Historical prior-source oracle: Existing exact unchanged-selector receipts remain the only current execution evidence. Changed source selectors have no fresh execution result. Historical prior-source oracle account: Existing exact unchanged-selector receipts remain the only current execution evidence. Changed source selectors have no fresh execution result. Historical prior-source oracle account: Fill the global watcher-child reservation, require native desktop, paired-runtime, and WSL setup to retain typed transient capacity and retry after one serialized release, then cancel pending desktop and paired-client watches and require no retry. Hold one live anchor while aborting 10,000 callers at the capacity, quarantine-assignment, runtime-root, relay-setup, and SSH-registration layers; require exactly one explicit waiter and one base-promise reaction to remain, then zero after settlement. Hold an old relay unsubscribe, queue one same-root setup leader and one anchor, then abort 10,000 joiners; require every cancellation to settle before teardown releases, exactly one anchor waiter, and only one later physical install. Advance the former paired-web callback deadline while capacity remains full, require cancellation ownership to stay live, publish the server token before ready, and require files.unwatch to abort setup before a later slot release can install a native root. Reject the first ready-watch unwatch with a physical-exit promise, resolve that exit later, and require a fresh physical unwatch attempt, registry release, and exactly one stream end. Fill one healthy shared child with multiple roots, map a standard SSH repository's base, Git common directory, and worktree roots to one relay supervisor, then inject IPC disconnect-before-exit, queued and active cancellation, operations during termination, kill-without-exit, repeated unkillable generations, initial and recovery timeouts, and shard failure. During local and SSH deletion, attempt a late terminal spawn whose worktree identity and cwd name different roots, enumerate a non-prefix remote PTY by authoritative worktree ownership, withhold daemon/relay exit callbacks, make attach positively prove a concurrent-shutdown PID dead, and make the provider unavailable; require every acquired admission to roll back on rejection, dead-PID proof to settle physical exit before reaping, timeout to retain ownership, and Git/filesystem deletion never to start. Require bounded watcher process count, physical-exit settlement, healthy-root recovery, stale-generation rejection, overflow plus later events, and fail-closed destructive cleanup. Verify relay build hashes and remote install probes require both relay.js and relay-watcher.js. Launch the built relay with a real PTY and Parcel watch, kill only the watcher child, and require the same relay PID and PTY to answer before a replacement child delivers overflow and a post-crash event.", + "oracle": "Audit3553/3559 current scope: delete worktrees-delete-pty-teardown.test.ts and worktrees-issue-command-overrides.test.ts entirely and remove three cases from worktrees-orphan-directory-cleanup.test.ts (6 authored/6 finite removals). Intentionally retire typed unserved-SSH create/delete admission, exact-name collision/error-token, proven-orphan no-force confirmation, standalone .git-directory refusal, and SSH-owned local-path preservation alarms. Retain only actual authorized orphan deletion and no-marker notes-file preservation; they are not equivalent replacements. The no-marker fixture resolves the default getRepos owner, so it does not prove temporary owner-repo binding. Git/Store/PTY/notifications are mocked; no physical process stop, durable metadata, real SSH delivery or native UI proof is claimed. Unchanged historical oracle, coverage, evidenceRuns and escaped receipts retain their original scopes only; no fresh execution or transferred pass validates the reduced sources. Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3460 final-v2 composes accepted AUTHOR3454/PEER3455 on actual post263 main: intentionally removes99 authored/106 potential expanded cases across16 selected drivers, including7 whole suites, retaining24 authored/26 potential cases. No equivalent replacement or fresh runtime is claimed. Broad root-unit discovery loses whole cross-client workspace-sync, terminal-preview, picker-skill-hint and close-routing suites plus reduced worker-stop, remote-owned-record, scanner, mobile-bundle and daemon-health samples. Separate E2E discovery also loses the paired mobile/desktop mirror journey and two unit drivers. Retained callback/table/outer-loop/guard bytes are exact; retained checks do not replace the removed alarms. No app, native device, mixed-version, remote or cross-platform execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The retained runtime registration driver loses three complete imported fragments: changed-working-label title facts and suppressed-frame tracked snapshots; refusing SSH GitLab MR-base lookup before local Git/glab calls; and missing folder-workspace path creation/update refusal before spawn/store mutation. These alarms are intentionally lost. Other imported fragments and separately rooted dated receipts are unchanged; they are not equivalent replacement coverage. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2090 (sources2088/2089): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2086 (sources2082/2083): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-handles-part-02.spec.ts: recovers a moved SSH pane through the tab it now sits in, not the one its lease froze; reports a plain expired orphan; does not report a pane whose only recent lease was superseded; does not report a pane whose only recent lease had its relay id recycled; picks the eligible orphan over a superseded predecessor that matches first; collects the same tabs the per-tab reader reports, in one sweep of the leases; reports no candidate when no lease names any of the worktree tabs. src/main/runtime/orca-runtime-tests/runtime-availability.spec.ts: starts unavailable with no authoritative window; reports the configured Windows terminal shell on status; reports floating workspace availability from settings on status; polls floating tabs with targeted PTY liveness and no repo/provider inventory; hydrates persisted tabs when the store cannot report repos; advertises browser screencast only when a renderer window is available; routes mobile Codex reset consumption through the account mutation coordinator; maps a definite pre-provider rejection into an authoritative current snapshot. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2056 (sources2054/2055): Intentionally retire these low-value alarms without replacement: src/renderer/src/runtime/files/watching/runtime-file-client-watch.test.ts: delivers remote file changes to listeners for the same worktree. src/main/runtime/orca-runtime.test.ts: ordinary registration cleanup removes only await import('./orca-runtime-tests/mobile-session-tabs-part-14.spec'). src/main/runtime/orca-runtime-tests/terminal-side-effect-facts-part-03.spec.ts: arms the stale-title timer for a seeded working title; arms the Command Code scrape from the noted spawn command; prefers the tracked title over a stale renderer lastTitle in the hydration seed; writes recent output into the current runtime output map; uses a later title handler replacement with the runtime receiver; calls the current injected listener with the runtime receiver; keeps optional reconciliation lazy after its callback is removed. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-14.spec.ts: lists them, so a paired client can mirror a migrated editor; closes one by retiring it from the host session; keeps an unsaved draft unless the close is forced; shares aggregate work but leaves a replacement pending promise alone; lets a targeted refresh finish while an aggregate refresh waits; refuses a close when the store lacks %s before the guard; keeps a later missing %s on the original caught failure path. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Exact retired assertionRefs, preserved as historical objects only: {\n \"file\": \"src/renderer/src/runtime/files/watching/runtime-file-client-watch.test.ts\",\n \"assertions\": [\n \"terminal watcher error and setup failure evict the shared entry before retrying callbacks run\",\n \"an old stream end cannot evict the replacement stream, which still delivers to its listener\",\n \"two same-worktree listeners share a synthetic subscription until the final stop unsubscribes and sends unwatch\"\n ]\n } Historical prior account (preserved verbatim): Audit1973: Delete the standalone mocked watcher cleanup suite and its entire sole-file prospective command. Retire the removed file-command transport-abort alias; preserve the independent cancellation suite and final-overflow/error/end cleanup callback, and13 unrelated receipt objects. Whole original changed dated receipts are retired, never narrowed or transferred. This metadata review claims no fresh execution, equivalent replacement, provider/PTY, live remote/mobile, rendered, cross-platform, soak or measured-budget result. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Partial runtime files-watch suite retains controlled Windows close-event waiting, shared-client survival, late physical-exit ownership/deadline cleanup and shutdown joining. Its combined Windows close/deadline/late-exit alias remains backed by unchanged callbacks. Retire native-error-close, pre-publication setup ownership and failed-delete rearm aliases; associated delegation, failure forwarding, removal routing and recovery examples are lost. These tests inject watch/subscription seams and do not observe an actual native watcher or physical process. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1925 (sources1922/1923): Partial runtime watcher-host suite loses mocked option/event/overflow/resubscription forwarding, supplied teardown-error propagation, aborted-signal/no-call and private ten-thousand-waiter-count inventories. Retire the 10,000-caller bound and failed-recovery-subscriber-ending assertion aliases. Keep same-root owner cancellation, pre-ready replacement, held root-close join, stale-generation recovery delivery and initial failed-root ownership mappings backed by retained real-host callbacks with injected subscription/error/physical-exit promises. No actual native watcher, process termination or resource bound was observed. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1683 retires changed-source assertion and execution scope without equivalent replacement coverage. Existing unchanged-selector receipts remain scoped to their original sources. Historical prior-source oracle: Existing exact unchanged-selector receipts remain the only current execution evidence. Changed source selectors have no fresh execution result. Historical prior-source oracle account: Existing exact unchanged-selector receipts remain the only current execution evidence. Changed source selectors have no fresh execution result. Historical prior-source oracle account: Fill the global watcher-child reservation, require native desktop, paired-runtime, and WSL setup to retain typed transient capacity and retry after one serialized release, then cancel pending desktop and paired-client watches and require no retry. Hold one live anchor while aborting 10,000 callers at the capacity, quarantine-assignment, runtime-root, relay-setup, and SSH-registration layers; require exactly one explicit waiter and one base-promise reaction to remain, then zero after settlement. Hold an old relay unsubscribe, queue one same-root setup leader and one anchor, then abort 10,000 joiners; require every cancellation to settle before teardown releases, exactly one anchor waiter, and only one later physical install. Advance the former paired-web callback deadline while capacity remains full, require cancellation ownership to stay live, publish the server token before ready, and require files.unwatch to abort setup before a later slot release can install a native root. Reject the first ready-watch unwatch with a physical-exit promise, resolve that exit later, and require a fresh physical unwatch attempt, registry release, and exactly one stream end. Fill one healthy shared child with multiple roots, map a standard SSH repository's base, Git common directory, and worktree roots to one relay supervisor, then inject IPC disconnect-before-exit, queued and active cancellation, operations during termination, kill-without-exit, repeated unkillable generations, initial and recovery timeouts, and shard failure. During local and SSH deletion, attempt a late terminal spawn whose worktree identity and cwd name different roots, enumerate a non-prefix remote PTY by authoritative worktree ownership, withhold daemon/relay exit callbacks, make attach positively prove a concurrent-shutdown PID dead, and make the provider unavailable; require every acquired admission to roll back on rejection, dead-PID proof to settle physical exit before reaping, timeout to retain ownership, and Git/filesystem deletion never to start. Require bounded watcher process count, physical-exit settlement, healthy-root recovery, stale-generation rejection, overflow plus later events, and fail-closed destructive cleanup. Verify relay build hashes and remote install probes require both relay.js and relay-watcher.js. Launch the built relay with a real PTY and Parcel watch, kill only the watcher child, and require the same relay PID and PTY to answer before a replacement child delivers overflow and a post-crash event.", "commands": [ - "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/filesystem/watchers/filesystem-watcher.test.ts src/main/ipc/filesystem/watcher-process/child-lifecycle/parcel-watcher-child-registry.test.ts src/main/ipc/filesystem/watcher-process/parcel-watcher-process.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-supervisor-capacity.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-disconnect-termination.test.ts src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts src/main/ipc/worktrees-delete-pty-teardown.test.ts src/main/ipc/worktrees-orphan-directory-cleanup.test.ts src/main/ipc/worktrees-remove-preflight.test.ts src/main/workspaces/removal/watcher-removal-gate.test.ts src/main/runtime/files/watch/orca-runtime-files-watch.test.ts src/main/runtime/workspaces/removal/worktree-teardown.test.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/methods/files/files.test.ts src/main/runtime/rpc/methods/files/files-watch-cancellation.test.ts src/renderer/src/web/web-runtime-client.test.ts src/shared/cross-platform-path.test.ts src/shared/physical-exit-tracker.test.ts src/shared/promise-settlement-waiters.test.ts src/main/ipc/pty/pty-spawn-cwd-fallback.test.ts src/main/providers/local-pty/local-pty-provider-shutdown.test.ts src/main/daemon/providers/adapter/testing/daemon-pty-adapter.test.ts src/main/daemon/pty-subprocess/testing/pty-subprocess.test.ts src/main/daemon/sessions/testing/session.test.ts src/main/daemon/terminal-host/testing/terminal-host.test.ts src/main/daemon/terminal-host/testing/terminal-host-session-reaping-leak.test.ts --reporter=dot", - "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/filesystem/watchers/filesystem-watcher.test.ts src/main/ipc/filesystem/watcher-process/child-lifecycle/parcel-watcher-child-registry.test.ts src/main/ipc/filesystem/watcher-process/parcel-watcher-process.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-supervisor-capacity.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-disconnect-termination.test.ts src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts src/main/ipc/worktrees-delete-pty-teardown.test.ts src/main/ipc/worktrees-orphan-directory-cleanup.test.ts src/main/ipc/worktrees-remove-preflight.test.ts src/main/workspaces/removal/watcher-removal-gate.test.ts src/main/runtime/files/watch/orca-runtime-files-watch.test.ts src/main/runtime/workspaces/removal/worktree-teardown.test.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/methods/files/files.test.ts src/main/runtime/rpc/methods/files/files-watch-cancellation.test.ts src/renderer/src/web/web-runtime-client.test.ts src/shared/cross-platform-path.test.ts src/shared/physical-exit-tracker.test.ts src/main/ipc/pty/pty-spawn-cwd-fallback.test.ts src/main/providers/local-pty/local-pty-provider-shutdown.test.ts src/main/daemon/providers/adapter/testing/daemon-pty-adapter.test.ts src/main/daemon/pty-subprocess/testing/pty-subprocess.test.ts src/main/daemon/sessions/testing/session.test.ts src/main/daemon/terminal-host/testing/terminal-host.test.ts src/main/daemon/terminal-host/testing/terminal-host-session-reaping-leak.test.ts --reporter=dot", - "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/filesystem/watchers/filesystem-watcher.test.ts src/main/ipc/filesystem/watcher-process/parcel-watcher-process.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-disconnect-termination.test.ts src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts src/main/ipc/worktrees-delete-pty-teardown.test.ts src/main/ipc/worktrees-orphan-directory-cleanup.test.ts src/main/ipc/worktrees-remove-preflight.test.ts src/main/workspaces/removal/watcher-removal-gate.test.ts src/main/runtime/files/watch/orca-runtime-files-watch.test.ts src/main/runtime/workspaces/removal/worktree-teardown.test.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/methods/files/files.test.ts src/renderer/src/web/web-runtime-client.test.ts src/shared/cross-platform-path.test.ts src/main/ipc/pty/pty-spawn-cwd-fallback.test.ts src/main/providers/local-pty/local-pty-provider-shutdown.test.ts src/main/daemon/pty-subprocess/testing/pty-subprocess.test.ts src/main/daemon/sessions/testing/session.test.ts src/main/daemon/terminal-host/testing/terminal-host.test.ts src/main/daemon/terminal-host/testing/terminal-host-session-reaping-leak.test.ts --reporter=dot", - "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/filesystem/watcher-process/parcel-watcher-process.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-disconnect-termination.test.ts src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts src/main/ipc/worktrees-delete-pty-teardown.test.ts src/main/ipc/worktrees-orphan-directory-cleanup.test.ts src/main/ipc/worktrees-remove-preflight.test.ts src/main/workspaces/removal/watcher-removal-gate.test.ts src/main/runtime/files/watch/orca-runtime-files-watch.test.ts src/main/runtime/workspaces/removal/worktree-teardown.test.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/methods/files/files.test.ts src/renderer/src/web/web-runtime-client.test.ts src/shared/cross-platform-path.test.ts src/main/ipc/pty/pty-spawn-cwd-fallback.test.ts src/main/providers/local-pty/local-pty-provider-shutdown.test.ts src/main/daemon/pty-subprocess/testing/pty-subprocess.test.ts src/main/daemon/sessions/testing/session.test.ts src/main/daemon/terminal-host/testing/terminal-host.test.ts src/main/daemon/terminal-host/testing/terminal-host-session-reaping-leak.test.ts", - "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/filesystem/watcher-process/parcel-watcher-process.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-disconnect-termination.test.ts src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts src/main/ipc/worktrees-delete-pty-teardown.test.ts src/main/ipc/worktrees-orphan-directory-cleanup.test.ts src/main/ipc/worktrees-remove-preflight.test.ts src/main/workspaces/removal/watcher-removal-gate.test.ts src/main/runtime/files/watch/orca-runtime-files-watch.test.ts src/main/runtime/workspaces/removal/worktree-teardown.test.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/methods/files/files.test.ts src/renderer/src/web/web-runtime-client.test.ts src/shared/cross-platform-path.test.ts src/main/ipc/pty/pty-spawn-cwd-fallback.test.ts src/main/providers/local-pty/local-pty-provider-shutdown.test.ts src/main/daemon/sessions/testing/session.test.ts src/main/daemon/terminal-host/testing/terminal-host.test.ts src/main/daemon/terminal-host/testing/terminal-host-session-reaping-leak.test.ts", - "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/filesystem/watcher-process/parcel-watcher-process.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-disconnect-termination.test.ts src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts src/main/ipc/worktrees-delete-pty-teardown.test.ts src/main/ipc/worktrees-orphan-directory-cleanup.test.ts src/main/ipc/worktrees-remove-preflight.test.ts src/main/workspaces/removal/watcher-removal-gate.test.ts src/main/runtime/files/watch/orca-runtime-files-watch.test.ts src/main/runtime/workspaces/removal/worktree-teardown.test.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/methods/files/files.test.ts src/renderer/src/web/web-runtime-client.test.ts src/shared/cross-platform-path.test.ts src/main/ipc/pty/pty-spawn-cwd-fallback.test.ts src/main/daemon/terminal-host/testing/terminal-host.test.ts src/main/daemon/terminal-host/testing/terminal-host-session-reaping-leak.test.ts", - "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/filesystem/watcher-process/parcel-watcher-process.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-disconnect-termination.test.ts src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts src/main/ipc/worktrees-delete-pty-teardown.test.ts src/main/ipc/worktrees-orphan-directory-cleanup.test.ts src/main/ipc/worktrees-remove-preflight.test.ts src/main/workspaces/removal/watcher-removal-gate.test.ts src/main/runtime/files/watch/orca-runtime-files-watch.test.ts src/main/runtime/workspaces/removal/worktree-teardown.test.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/methods/files/files.test.ts src/renderer/src/web/web-runtime-client.test.ts src/shared/cross-platform-path.test.ts", + "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/filesystem/watchers/filesystem-watcher.test.ts src/main/ipc/filesystem/watcher-process/child-lifecycle/parcel-watcher-child-registry.test.ts src/main/ipc/filesystem/watcher-process/parcel-watcher-process.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-supervisor-capacity.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-disconnect-termination.test.ts src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts src/main/ipc/worktrees-orphan-directory-cleanup.test.ts src/main/ipc/worktrees-remove-preflight.test.ts src/main/workspaces/removal/watcher-removal-gate.test.ts src/main/runtime/files/watch/orca-runtime-files-watch.test.ts src/main/runtime/workspaces/removal/worktree-teardown.test.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/methods/files/files.test.ts src/main/runtime/rpc/methods/files/files-watch-cancellation.test.ts src/renderer/src/web/web-runtime-client.test.ts src/shared/cross-platform-path.test.ts src/shared/physical-exit-tracker.test.ts src/shared/promise-settlement-waiters.test.ts src/main/ipc/pty/pty-spawn-cwd-fallback.test.ts src/main/providers/local-pty/local-pty-provider-shutdown.test.ts src/main/daemon/providers/adapter/testing/daemon-pty-adapter.test.ts src/main/daemon/pty-subprocess/testing/pty-subprocess.test.ts src/main/daemon/sessions/testing/session.test.ts src/main/daemon/terminal-host/testing/terminal-host.test.ts src/main/daemon/terminal-host/testing/terminal-host-session-reaping-leak.test.ts --reporter=dot", + "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/filesystem/watchers/filesystem-watcher.test.ts src/main/ipc/filesystem/watcher-process/child-lifecycle/parcel-watcher-child-registry.test.ts src/main/ipc/filesystem/watcher-process/parcel-watcher-process.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-supervisor-capacity.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-disconnect-termination.test.ts src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts src/main/ipc/worktrees-orphan-directory-cleanup.test.ts src/main/ipc/worktrees-remove-preflight.test.ts src/main/workspaces/removal/watcher-removal-gate.test.ts src/main/runtime/files/watch/orca-runtime-files-watch.test.ts src/main/runtime/workspaces/removal/worktree-teardown.test.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/methods/files/files.test.ts src/main/runtime/rpc/methods/files/files-watch-cancellation.test.ts src/renderer/src/web/web-runtime-client.test.ts src/shared/cross-platform-path.test.ts src/shared/physical-exit-tracker.test.ts src/main/ipc/pty/pty-spawn-cwd-fallback.test.ts src/main/providers/local-pty/local-pty-provider-shutdown.test.ts src/main/daemon/providers/adapter/testing/daemon-pty-adapter.test.ts src/main/daemon/pty-subprocess/testing/pty-subprocess.test.ts src/main/daemon/sessions/testing/session.test.ts src/main/daemon/terminal-host/testing/terminal-host.test.ts src/main/daemon/terminal-host/testing/terminal-host-session-reaping-leak.test.ts --reporter=dot", + "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/filesystem/watchers/filesystem-watcher.test.ts src/main/ipc/filesystem/watcher-process/parcel-watcher-process.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-disconnect-termination.test.ts src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts src/main/ipc/worktrees-orphan-directory-cleanup.test.ts src/main/ipc/worktrees-remove-preflight.test.ts src/main/workspaces/removal/watcher-removal-gate.test.ts src/main/runtime/files/watch/orca-runtime-files-watch.test.ts src/main/runtime/workspaces/removal/worktree-teardown.test.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/methods/files/files.test.ts src/renderer/src/web/web-runtime-client.test.ts src/shared/cross-platform-path.test.ts src/main/ipc/pty/pty-spawn-cwd-fallback.test.ts src/main/providers/local-pty/local-pty-provider-shutdown.test.ts src/main/daemon/pty-subprocess/testing/pty-subprocess.test.ts src/main/daemon/sessions/testing/session.test.ts src/main/daemon/terminal-host/testing/terminal-host.test.ts src/main/daemon/terminal-host/testing/terminal-host-session-reaping-leak.test.ts --reporter=dot", + "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/filesystem/watcher-process/parcel-watcher-process.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-disconnect-termination.test.ts src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts src/main/ipc/worktrees-orphan-directory-cleanup.test.ts src/main/ipc/worktrees-remove-preflight.test.ts src/main/workspaces/removal/watcher-removal-gate.test.ts src/main/runtime/files/watch/orca-runtime-files-watch.test.ts src/main/runtime/workspaces/removal/worktree-teardown.test.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/methods/files/files.test.ts src/renderer/src/web/web-runtime-client.test.ts src/shared/cross-platform-path.test.ts src/main/ipc/pty/pty-spawn-cwd-fallback.test.ts src/main/providers/local-pty/local-pty-provider-shutdown.test.ts src/main/daemon/pty-subprocess/testing/pty-subprocess.test.ts src/main/daemon/sessions/testing/session.test.ts src/main/daemon/terminal-host/testing/terminal-host.test.ts src/main/daemon/terminal-host/testing/terminal-host-session-reaping-leak.test.ts", + "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/filesystem/watcher-process/parcel-watcher-process.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-disconnect-termination.test.ts src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts src/main/ipc/worktrees-orphan-directory-cleanup.test.ts src/main/ipc/worktrees-remove-preflight.test.ts src/main/workspaces/removal/watcher-removal-gate.test.ts src/main/runtime/files/watch/orca-runtime-files-watch.test.ts src/main/runtime/workspaces/removal/worktree-teardown.test.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/methods/files/files.test.ts src/renderer/src/web/web-runtime-client.test.ts src/shared/cross-platform-path.test.ts src/main/ipc/pty/pty-spawn-cwd-fallback.test.ts src/main/providers/local-pty/local-pty-provider-shutdown.test.ts src/main/daemon/sessions/testing/session.test.ts src/main/daemon/terminal-host/testing/terminal-host.test.ts src/main/daemon/terminal-host/testing/terminal-host-session-reaping-leak.test.ts", + "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/filesystem/watcher-process/parcel-watcher-process.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-disconnect-termination.test.ts src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts src/main/ipc/worktrees-orphan-directory-cleanup.test.ts src/main/ipc/worktrees-remove-preflight.test.ts src/main/workspaces/removal/watcher-removal-gate.test.ts src/main/runtime/files/watch/orca-runtime-files-watch.test.ts src/main/runtime/workspaces/removal/worktree-teardown.test.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/methods/files/files.test.ts src/renderer/src/web/web-runtime-client.test.ts src/shared/cross-platform-path.test.ts src/main/ipc/pty/pty-spawn-cwd-fallback.test.ts src/main/daemon/terminal-host/testing/terminal-host.test.ts src/main/daemon/terminal-host/testing/terminal-host-session-reaping-leak.test.ts", + "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/filesystem/watcher-process/parcel-watcher-process.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-disconnect-termination.test.ts src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts src/main/ipc/worktrees-orphan-directory-cleanup.test.ts src/main/ipc/worktrees-remove-preflight.test.ts src/main/workspaces/removal/watcher-removal-gate.test.ts src/main/runtime/files/watch/orca-runtime-files-watch.test.ts src/main/runtime/workspaces/removal/worktree-teardown.test.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/methods/files/files.test.ts src/renderer/src/web/web-runtime-client.test.ts src/shared/cross-platform-path.test.ts", "pnpm run build:electron-vite", "node config/scripts/diagnostics/runtime-watchers/runtime-file-watcher-resource-probe.mjs", "node config/scripts/diagnostics/runtime-watchers/runtime-file-watcher-fault-harness.mjs", @@ -12361,7 +12361,6 @@ "src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts", "src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts", "src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts", - "src/main/ipc/worktrees-delete-pty-teardown.test.ts", "src/main/ipc/worktrees-orphan-directory-cleanup.test.ts", "src/main/ipc/worktrees-remove-preflight.test.ts", "src/main/workspaces/removal/watcher-removal-gate.test.ts", @@ -12658,6 +12657,7 @@ "Add a packaged orca serve pairing scenario that proves a connected client refreshes after child replacement." ], "knownGaps": [ + "Audit3553/3559 current scope: delete worktrees-delete-pty-teardown.test.ts and worktrees-issue-command-overrides.test.ts entirely and remove three cases from worktrees-orphan-directory-cleanup.test.ts (6 authored/6 finite removals). Intentionally retire typed unserved-SSH create/delete admission, exact-name collision/error-token, proven-orphan no-force confirmation, standalone .git-directory refusal, and SSH-owned local-path preservation alarms. Retain only actual authorized orphan deletion and no-marker notes-file preservation; they are not equivalent replacements. The no-marker fixture resolves the default getRepos owner, so it does not prove temporary owner-repo binding. Git/Store/PTY/notifications are mocked; no physical process stop, durable metadata, real SSH delivery or native UI proof is claimed. Unchanged historical oracle, coverage, evidenceRuns and escaped receipts retain their original scopes only; no fresh execution or transferred pass validates the reduced sources.", "Launch direct-owner cleanup current scope: This gate’s unfiltered unit command now selects a suite reduced by 82 whole registrations / 159 potential table examples in 10 whole test Programs, including 8 registrations / 15 examples from five once-registered create-message factory closures. Lost alarms include frozen first-message/options and legacy-host fallback; distinct actions and redelivery; uncertain/refused/reloaded input custody, Retry and late proof; rich prompt/attachment upload Stop/close and handback; ordinary draft/editor/IME/StrictMode lifetime and accepted topology; provisional fallback tab placement/group/pin/order and rich worktree handoff. Four whole consent/resume-identity cases remain in two Programs; they do not replace these alarms. This entry governs current reduced-unit scope and qualifies any broader implication of unchanged oracle/coverageNotes history. Entire dated evidenceRuns, original commands/currentCommand values, counts/results and existing history retain their original scopes only; no fresh execution, transferred pass, replacement equivalence or prior upstream-removal credit is claimed.", "Audit2145 accepts removal of canned runtime metadata/guard/forwarding, fake in-memory persistence, private work-count and wrapper-duplication tests from the construction/orca-runtime.test.ts aggregate, plus the two automation run-context guards from runtime-workspace-launch-contracts.test.ts. Retained race, framing, physical completion and ownership callbacks remain byte-identical. Existing commands remain valid selectors; historical receipts stay unchanged and provide no fresh validation or replacement coverage for the reduced suites.", "Audit2133 current-parent preparation (988 authored removals; independently reviewed2099/2107/2115/2116 decisions; upstream clear-buffer and recovery-image regressions retained): Intentionally retire these low-value alarms without replacement: src/main/runtime/construction/orca-runtime.test.ts: remove only literal registration import/call pairs await import('../orca-runtime-tests/worktree-setup-and-startup.spec'), await import('../orca-runtime-tests/gitlab-and-pr-bases.spec'); no additional authored credit. src/main/runtime/orca-runtime-tests/targeting-and-resilience.spec.ts: routes tab switch and capture start by explicit page id; accepts focus on tab switch without altering bridge args (focus is main-side concern); keeps terminal cwd resolution fail-soft when the provider is unavailable. src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-04.spec.ts: drops the bounded scan cache when orphan-cleanup removal completes; clears optimistic reconcile tokens when a CLI worktree removal succeeds; emits drift without mutating when the fetched base fast-forwards created HEAD; emits base_changed without mutation when the fetched base rewrote history; skips stale-token reconciles without mutating or emitting stale status; emits unknown without mutation when fetch fails or the base ref is missing. src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-03.spec.ts: routes runtime preserved-branch force-delete through the selected WSL project runtime; replies to a waiting client with the error of a delete that fails after acceptance; treats forced runtime deletion of an already-missing unregistered worktree as cleanup; treats normal runtime deletion of an already-missing unregistered worktree as cleanup. src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-02.spec.ts: warns that a missing-repo removal only forgot the workspace; restores runtime watchers when CLI worktree deletion fails after teardown; does not re-spend the drain budget on a removal after a wedged install was abandoned; routes runtime worktree removal through the selected WSL project runtime; force-deletes a branch that was preserved by runtime worktree removal. src/main/runtime/orca-runtime-tests/gitlab-and-pr-bases-part-02.spec.ts: records GitLab pasted-project recents only after successful runtime lookup; threads explicit origin preference through runtime WSL PR base resolution; resolves local GitLab fork MR bases from the target project MR head ref; captures the fork MR head from a dedicated ref, not the shared FETCH_HEAD; keeps the durable MR head when the head fetch fails but the local ref resolves; routes runtime GitLab fork MR base git calls through the selected WSL project runtime. src/main/runtime/orca-runtime-tests/gitlab-and-pr-bases.spec.ts: passes SSH connection ids through GitLab task operations; routes runtime GitLab issue, MR, work-item, and todo actions through the selected WSL project runtime; routes runtime GitLab MR details, review-management, job, and pasted URL actions through the selected WSL project runtime; normalizes runtime GitLab issue list arguments like the desktop IPC path. src/main/runtime/orca-runtime-tests/worktree-setup-and-startup-part-03.spec.ts: does not warn when setup is explicitly skipped for CLI-created worktrees; puts the agent a create starts in the first default tab, as the window does; keeps the 8s main-runtime startup readiness budget for agents without an override; gives OpenCode the 20s main-runtime startup readiness budget; launches explicit startup agents with prompts for CLI-created worktrees. src/main/runtime/orca-runtime-tests/worktree-setup-and-startup.spec.ts: returns a setup launch payload for CLI-created worktrees when hooks are explicitly enabled; does not hand the window a setup main never tried to start; passes the selected Windows setup shell into runtime runner generation; follows normal setup policy for CLI-created worktrees without activating them; uses returned WSL setup shell metadata when runtime spawns setup; uses the shell-aware setup runner for windowless creates without a startup command; reports the in-process setup hook as running when nothing can launch the runner. src/main/runtime/orca-runtime-tests/terminal-sleep-and-teardown-part-02.spec.ts: coalesces two clients sleeping the same host worktree; stops exactly the expected live PTYs for a worktree; explains a refused workspace close instead of throwing the raw refusal enum. src/main/runtime/orca-runtime-tests/mobile-summaries-part-03.spec.ts: marks the desktop-active worktree as isActive; includes SSH-backed worktrees in the mobile worktree summary; caches repeated malformed path misses before normalizing them again. src/main/runtime/orca-runtime-tests/mobile-summaries-part-02.spec.ts: keeps title-only foreground work ahead of monitoring in another pane; keeps a local hook row while a connected PTY still backs its pane; keeps an OSC row via its connected PTY after the pane binding is cleared. src/main/runtime/orca-runtime-tests/mobile-summaries.spec.ts: resolves files through a source-visible linked-checkout worktree; marks saved session tabs with live PTYs as host sidebar activity; attributes live legacy PTYs from saved layout bindings when their panes are hidden; marks saved browser tabs as host sidebar activity like desktop; falls back to the path-keyed GitHub cache entry; includes folder workspaces in compact worktree summaries for mobile; attaches inline agent rows from the latest OSC 9999 status; projects monitoring over a title-derived working status. src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-03.spec.ts: pushes to an idle pane when the only live waiter filters out the message type; resolves a registered waiter without PTY-injecting when the leaf is already idle; keeps already-idle status after tui-idle wait for immediate message delivery; resolves message waiters when notifyMessageArrived is called; does not resolve type-filtered message waiters for unrelated message types; removes message waiter abort listeners after message arrival; resolves message waiters on timeout when no message arrives. src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-02.spec.ts: delivers pending mail via notifyMessageArrived when the recipient is already idle; does not keep retrying when every pending row was already pointed; repoints pending rows restored with a live-idle coordinator. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-13.spec.ts: rolls back a half-created terminal whose surface never publishes. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-12.spec.ts: rejects startup prompts for agents that require post-ready stdin; uses portable Unix quoting for mobile agent launch commands in WSL project runtimes; keeps PowerShell quoting for mobile agent launch commands in Windows host runtimes; uses cmd.exe quoting for mobile agent launch commands in local Windows host runtimes; publishes headless mobile session agent identity with synthesized PTY status; rejects disabled mobile session agent launches before spawning; validates mobile terminal insertion anchors before resolving agent launch commands. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-11.spec.ts: coalesces a reconnect close burst onto one authoritative PTY inventory; keeps an explicit user close destructive while the PTY is live; builds mobile session agent launch commands on the runtime host; injects mobile quick-command prompts into the host-built agent startup command. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-10.spec.ts: closes persisted headless terminal parents before any prior list call; tears down a serve-owned headless tab on close while a renderer is attached so it cannot resurrect; defers a renderer-published pending tab to the renderer instead of tearing it down; returns a delegated close outcome with no selection tombstone when the renderer owns the outcome; tears down a runtime pending shell the renderer never adopted on close. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-08.spec.ts: materializes a plain shell when the pending tab has no launch agent; hydrates legacy persisted terminal tabs without layout entries; reads and indexes the full agent-status snapshot once per mobile projection; kills persisted SSH PTYs when closing hydrated headless tabs before pane metadata is restored; materializes hydrated pending headless terminals with the persisted session identity. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-07.spec.ts: hydrates a persisted SSH-owned pane before an attached renderer publishes its graph. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-05.spec.ts: publishes the desktop's tab name over the live agent title, and the live title without one; uses fresh neutral PTY titles over stale mobile snapshot and OSC titles; operates PTY-backed mobile session terminals without a renderer graph; reuses pane close for live PTYs that do not own a renderer tab; lists PTY-backed mobile session terminals without a renderer graph; shows and resolves active PTY-backed mobile session terminals without a renderer graph. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-04.spec.ts: creates mobile session terminals for folder workspaces in a headless runtime server; spawns fresh headless SSH mobile session terminals instead of reattaching synthetic local ids; hydrates headless mobile session terminals from the host workspace session; keeps live headless mobile session terminals when a desktop renderer publishes without them; publishes laptop-created remote runtime split terminals to phone session tabs. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-02.spec.ts: does not report headless wake degradation without sleeping records; does not wake slept agents for non-mobile session-only activation; does not rewrite unread metadata when a mobile activation finds the worktree already read; returns unread:false from worktree.ps after a mobile activation clears the flag; materializes pending mobile session terminals without focusing desktop clients; closes browser mobile session tabs when addressed by browser workspace id. src/main/runtime/orca-runtime-tests/mobile-session-tabs.spec.ts: resolves mobile terminal surfaces by exact split leaf; closes the matching mobile terminal UUID leaf without closing the whole tab; closes the whole mobile terminal tab when addressed by parent tab id; activates the active split leaf when addressed by parent tab id; clears unread metadata on mobile worktree activation without focusing desktop clients; wakes slept agents on the host renderer when a phone activates a worktree. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-05.spec.ts: publishes the hook provider session on a headless mobile tab so native chat can address the transcript; recovers the agent type from the hook row when the pane was launched without an agent hint; reads one agent-status snapshot per projection, not one per terminal tab; waits for unknown-launch foreground owner before publishing Pi-compatible mobile status; keeps Pi-compatible mobile session status as Pi for an unknown-launch foreground pi PTY; keeps renderer-vetted mobile agent status for custom-titled terminals. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-04.spec.ts: publishes terminal surface agent status for paired web clients; preserves authoritative OMP identity for Pi-compatible remote terminal snapshots; normalizes a remote OMP title without republishing omitted launch identity; skips the foreground-process probe when the PTY launch agent is already known; probes the foreground process only on a status transition for unknown launch agents; passes the client signal and the slept-worktree refusal to the spawn. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-03.spec.ts: does not use stale runtime-created PTY status when a neutral PTY title exists; recognizes runtime-created Codex PTY handles from the ready prompt; recognizes runtime-created Antigravity PTY handles from the ready prompt; recognizes live leaf Antigravity terminals from the ready prompt; recognizes a later Antigravity header with a prompt but no model line; keeps mobile terminal surfaces visible while their leaf handle is pending. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-02.spec.ts: lets Claude agents management titles clear stale runtime-created title status; does not recognize live Claude agents panes from a Claude foreground process; does not let stale tab-level agent titles override current neutral pane titles; lists live terminals with fresh pane titles over stale tab titles; lets adopted neutral pane titles use non-shell foreground fallback; lets adopted neutral pane titles retry wrapper foregrounds until recognized; authorizes settled CLI prompts only after positive %s foreground identity; does not recognize arbitrary foreground TUIs as running agents; lets live neutral pane titles retry wrapper foregrounds until recognized; lets current Claude agents PTY titles override stale runtime-created OSC titles; does not let stale Claude agents OSC titles suppress current PTY title activity. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status.spec.ts: keeps preallocated terminal handles valid across renderer reloads; preserves runtime-created PTY process identity after graph unavailable; keeps prompt bindings fenced across runtime restarts without provider incarnation; recognizes runtime-created PTY handles with agent launch titles; does not treat a bare Cursor Agent native title as a running agent session; uses stale runtime-created PTY status when there is no title or foreground evidence. src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-07.spec.ts: keeps orphaned list and show writability aligned with the send gate. src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-04.spec.ts: cancels a coalesced SSH worker recovery retry when its provider disconnects. src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-03.spec.ts: retries renderer reveal before clearing an adopted legacy worker sleeping record. src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-02.spec.ts: recovers exported ORCA_TERMINAL_HANDLE from discovered live PTY sessions. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-12.spec.ts: applies ANSI erase-from-start line controls in retained previews; applies ANSI stripping for private or intermediate CSI line controls; applies ANSI stripping for unsupported erase-line modes; reveals an agent terminal under the pane the caller reserved; keeps a fresh %s pane host-owned when a desktop window is available; keeps desktop publication %s separate from the reserved host spawn; mints its own pane when the reserved leaf is malformed, and reports the one it used; starts an agent terminal with the model the caller picked; starts an agent terminal without a model flag when no pick was made; attributes a startup-agent launch named %s as %s; attributes a fresh agent session the host builds as unknown; leaves a bare agent command the user typed out of launch attribution. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-11.spec.ts: trims oversized terminal output bursts without per-line array shifts; builds terminal previews without mapping the full retained tail; keeps recent PTY replay output capped without needing previous data for large chunks; applies terminal redraw controls before retaining previews; applies ANSI terminal redraw controls before retaining previews; retains multi-line ANSI footer redraws instead of appending old frames; keeps the cursor column when erasing full lines in multi-line redraws; applies ANSI split redraw controls without leaking raw params. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-10.spec.ts: returns renderer visible screen lines through terminal.read RPC JSON result; separates composer drafts from provider-owned terminal screens. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-09.spec.ts: reports an in-flight PTY focus as not navigated when its notifier disappears; reports graph-leaf focus as not navigated without a host notifier; grounds the provider mode tracker before a later chunk, in the emulator order; grounds an in-flight provider snapshot capture so it cannot publish the pre-reset screen; waits for terminal exit and resolves with the exit status; keeps terminal read payloads bounded while retained output remains pageable; trims terminal read preview character budget without per-line array shifts. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-08.spec.ts: submits a silent Claude composer once after the bounded render fallback; chunks large terminal.send text before provider writes; yields chunked terminal input through immediates between writes; yields while validating accepted large terminal.send text before provider writes; reveals a background terminal session when focusing its handle; replays captured launch config when focusing a background agent session; reveals background terminal sessions with the freshest PTY title; renames background terminal handles without requiring a visible tab; keeps a background terminal handle stable while reveal adoption is racing. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-07.spec.ts: resolves tui-idle when a stale Codex prompt is followed by Antigravity readiness; returns an agent-neutral blocked wait result for update prompts; returns an agent-neutral blocked wait result for workspace trust prompts; returns an agent-neutral blocked wait result for cwd selection prompts; returns a blocked wait result for Codex model migration prompts; returns a blocked wait result for Codex startup hook review prompts; returns an agent-neutral blocked wait result for generic interactive prompts; splits text and enter writes for background terminal handles; types the lead line for %s: %s; submits through the agent-specific PTY timing policy for %s. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-06.spec.ts: waits for exit on background terminal handles; keeps retained PTY transcript memory when controller refresh fails; resolves tui-idle from an Antigravity ready prompt preview; resolves Antigravity ready prompts with newline-heavy pasted tails without splitting; resolves live-leaf tui-idle from an Antigravity ready prompt preview. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-03.spec.ts: spawns the available Qoder command for a captured same-session resume: %s; does not use the local Windows shell setting for remote Windows bare agent creates; matches canonical bare agent commands when a command override is configured; keeps non-bare agent command terminal creates unchanged; keeps disabled bare agent command terminal creates unchanged; injects runtime hook receiver env into terminal sessions; creates background terminal sessions for a $label; creates background terminal sessions for a $label; enables Claude Agent Teams only for direct Claude launches when configured in-process; reveals Claude Agent Teams launches with the rewritten launch config; preserves Claude Agent Teams for sequenced Claude launches. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness.spec.ts: creates visible terminal sessions without asking the renderer to focus a tab; asks the pty controller for the requested shell instead of a startup command; knows the agent a flagged launch command names before any process read; retires a flagged command pane identity when the command exits; retires only receipted restored PTY authority on command completion and exit; forgets synthetic handles when disconnected PTY records are pruned. src/main/runtime/orca-runtime-tests/headless-snapshots.spec.ts: returns OSC titles from headless main terminal snapshots; adopts renderer-seeded titles into headless main terminal snapshots; returns cwd metadata from seeded headless main terminal snapshots; adopts OSC7 host metadata from seeded headless terminal scrollback; falls back to the renderer snapshot for hidden-output recovery without headless state; keeps an empty headless snapshot authoritative for hidden-output recovery; advances the absolute output sequence across a daemon stream gap; emits explicit OSC 9999 agent status from runtime PTY data. src/main/runtime/orca-runtime-tests/terminal-side-effect-facts-part-02.spec.ts: falls back to the provider snapshot for a restored PTY with no mounted renderer; falls back to the available mirror when authoritative provider history is unavailable; keeps an empty renderer snapshot when the provider has no retained content; keeps restored provider history authoritative after later live output; uses provider alternate-screen state while a partial model is unsafe; prefers the tracked title over the headless emulator lastTitle; keeps the cursor-agent literal in record-fallback snapshots only without a tracker title; touches mobile snapshots once for decorative spinner ticks, again on idle. src/main/runtime/orca-runtime-tests/local-worktree-creation-part-02.spec.ts: creates a same-repo PR branch override from a resolved head SHA and matching push target; skips broad remote fetch for an existing full-SHA PR base; creates a selected Bitbucket PR branch override from a matching remote branch; suffixes an existing PR when a matching push target lacks selected PR metadata; suffixes a matching push target branch when selected PR metadata has no PR number; suffixes a matching push target branch when the existing PR is different; suffixes a selected PR remote conflict when the PR lookup fails; checks out an unused runtime PR branch only when it is at the resolved head SHA; suffixes only the runtime worktree path when an exact PR branch checkout path exists. src/main/runtime/orca-runtime-tests/local-worktree-creation.spec.ts: returns runtime local base update suggestions from addWorktree; creates a runtime local worktree from a usable persisted local branch base; does not create a runtime local worktree when the refresh fails and no local base ref exists; creates a branchNameOverride worktree from the selected matching remote base ref. src/main/persistence/worktree-sort-order-write.test.ts: re-ranks every writable row when a blocked row sits in the requested order. src/main/runtime/integrations/skill/orca-runtime-skill-recovery.test.ts: continues skill management after startup recovery fails. src/main/runtime/orchestration/mailbox/routing/orchestration-mailbox-detached-routing.spec.ts: uses bounded ownership lookup for direct arrivals while the agent cannot take input; wakes a bound Run waiter when same-Run direct mail arrives; keeps an unowned detached-recipient message in its direct mailbox; uses primary-key lookups for bounded routing updates. src/main/runtime/orchestration/mailbox/terminal-pointer/orchestration-mailbox-delivery.test.ts: remove only literal registration import/call pairs import './orchestration-mailbox-pointer-cli-command.spec'; no additional authored credit. src/main/runtime/orchestration/mailbox/terminal-pointer/orchestration-mailbox-pointer-cli-command.spec.ts: renders the %s CLI command in a mailbox pointer. src/main/runtime/rpc/methods/terminal/terminal-multiplex-link-estimate.test.ts: sizes a slow high-latency link near twice its bandwidth-delay product; never paces a LAN, even when its client parses slowly; never paces a fast link; returns to legacy windows when the path RTT jumps. src/main/runtime/rpc/methods/terminal/terminal-snapshot-scrollback-rows.test.ts: steps down the row ladder when history is over the byte budget; tries the larger first row count, then the usual ladder within the same byte budget. src/main/runtime/rpc/server/pairing/network/windows-default-route-interfaces.test.ts: returns unknown when route inspection fails or emits malformed data; does not inspect routes outside Windows. src/main/runtime/rpc/server/pairing/pairing-offer-owner-cli.test.ts: is refused to a paired %s client; reports a failed widening instead of advertising an unreachable address. src/main/runtime/rpc/terminal-multiplex-end-verdict.test.ts: reports exited only when the owning runtime completes the exit waiter; reports unverifiable when the runtime cannot observe the exit waiter; reports unverifiable when a disconnected PTY resolves with unknown liveness. src/main/runtime/session-graph/headless-runtime-graph.spec.ts: lets a session tab inventory settle on a host no renderer publishes for. src/main/runtime/session-graph/runtime-graph-standing-contracts.test.ts: remove only literal registration import/call pairs import { registerHeadlessRuntimeGraphTests } from './headless-runtime-graph.spec', registerHeadlessRuntimeGraphTests(); no additional authored credit. src/main/runtime/session-tabs/structured-conversation-tab-title.test.ts: bumps the stored snapshot version only when the title changes. src/main/runtime/session-tabs/terminal-close-observed-exit.test.ts: keeps the healthy aggregate verification and delayed physical exit behavior; uses a stamped exit for the incarnation that was actually being closed. src/main/runtime/terminal/identity/terminal-identity-probe.test.ts: answers live for a structured worker without touching the PTY graph; answers live for a PTY handle the runtime still holds; runs the full leaf check for a handle with no live PTY; answers not-live for a stale handle; propagates \"could not look\" rather than reporting it as a dead handle. src/main/runtime/terminal/layout/paired-desktop-server-writes-layout-owner.test.ts: names the tab in main’s layout through terminal.rename; takes a window-style layout.apply close as already stopped, so the terminal runs on. src/main/runtime/terminal/lifecycle/pty-lifecycle-generation-retention.test.ts: retains no lifecycle generation after a spawn/exit cycle. src/main/runtime/terminal/lifecycle/run-terminal-user-open.test.ts: reads a freshly spawned terminal as used once a user opens its tab, not on a probe; keeps the open mark on the process a user open starts. src/main/runtime/terminal/surface/terminal-surface-replies-window.test.ts: from a desktop window that mounts both panes; from a server with no window. src/main/runtime/terminal/surface/workspace-layout-projection.test.ts: carries views for a partition that holds none yet. src/main/runtime/terminal/view/terminal-subscribe-exit-waiter-leak.test.ts: releases a signalled exit-waiter when the signal aborts. src/main/runtime/terminal/wait/orca-runtime-exact-worker-provider-session.test.ts: selects a local hook status for a local pane. src/main/runtime/terminal/wait/prime-agent-screen-readiness-transcripts.test.ts: keeps the footer and the bare caret on screen mid-turn; %s: the screen alone reads ready during %s. src/main/runtime/workspaces/startup/runtime-worktree-agent-startup.test.ts: uses per-launch arguments and preserves launch telemetry; attributes a startup agent whose caller named no surface as unknown; launches %s bare for an absent or blank prompt; attributes a %s draft launch named %s as %s; merges an automation's extras over the host defaults; threads worktree-create extras into the startup build; returns an enabled requested agent without detecting; returns the default when nothing was requested; detects instead of using the default when the requested agent is disabled; starts no agent when the default is blank; detects an enabled agent on this host when nothing usable was named; starts no agent when detection finds none. src/main/runtime/workspaces/startup/runtime-worktree-startup-readiness.test.ts: pastes and submits through composer readiness, never typing raw input; types the follow-up on WSL or SSH-to-Windows once the agent is in front, as before; cleans up the deadline when the composer was already captured; a draft that is not submitted is pasted on the box cursor, as before; takes the box after the grace when the row never comes, whatever repaints; takes the box at the deadline when the grace would outlast the budget; lets a dialog on screen veto a grace; refuses the deadline inside a pending grace while a shell is in front; takes the box at the deadline inside a pending grace with the agent in front; takes the box at the deadline after a refused grace once the agent is in front. src/main/skills/skill-install-transaction.test.ts: installs, previews unchanged bytes, and updates a clean receipt-owned skill; preserves the old version after a %s filesystem failure. src/main/speech/model-manager-download-resume.test.ts: requests the remaining bytes when a clean range response ends before the file total; bounds a server that advances by pathologically tiny segments forever. src/main/ssh/config/ssh-config-parser.test.ts: parses a basic host block; parses multiple host blocks; skips wildcard-only Host entries; skips Host entries with only pattern characters; parses IdentityFile with ~ expansion; parses Windows-style IdentityFile with ~ expansion; parses quoted scalar values like OpenSSH; parses equals-form scalar directives; parses IdentityAgent with ~ expansion; parses IdentitiesOnly; parses GSSAPIAuthentication; keeps the first GSSAPIAuthentication value like OpenSSH; parses ProxyCommand, ProxyUseFdpass, and ProxyJump; ignores comments and blank lines; handles case-insensitive keywords; stops current block on Match directive; returns empty array for empty input; creates one parsed host per concrete alias on a multi-pattern Host line; applies identity agent settings to every concrete alias on a multi-pattern Host line; defaults port to 22 for invalid port values; converts hosts to SshTarget objects; uses host alias as hostname when HostName is missing; skips hosts that are already imported; defaults username to empty string when not specified; carries through identityFile, identityAgent, identitiesOnly, proxyCommand, and jumpHost; carries through gssapiAuthentication; imports duplicate aliases only once and keeps the first concrete host; maps hosts for the picker and flags aliases already in Orca; de-dupes aliases and defaults missing fields; parses hostname, user, port from ssh -G output; collects multiple identity files; parses forwardagent yes; defaults port to 22 when missing; returns empty hostname when missing; returns undefined user when missing; handles empty output; skips lines without spaces; parses proxycommand and filters \"none\"; parses proxyusefdpass yes; parses proxyjump and filters \"none\"; handles ~ expansion in identity file paths; handles Windows-style ~ expansion in identity file paths; parses identityagent with ~ expansion; preserves identityagent none so auth can disable agent fallback; parses identitiesonly yes; parses gssapiauthentication; parses controlmaster options and filters controlpath none. src/main/ssh/config/ssh-known-hosts-source.test.ts: splits the space-separated list ssh -G prints on one line; expands ~ in a reported path; keeps a double-quoted path containing spaces whole; falls back to the default files when ssh -G reported nothing; drops an explicit none without falling back to the defaults; splits both known_hosts lists on one line each; reads StrictHostKeyChecking, including accept-new; reads HostKeyAlias and prefers it for the lookup; reads the non-default port; reports that the lookup came from an alias, so the port is not appended; reports no alias when the dialed host is used; unions the entries of every file; skips a path that is a directory; returns nothing when no file can be read; reports nothing unreadable when every file parses; does not count an empty file as unreadable; prefers HostKeyAlias over the resolved hostname; uses the dialed host, because a resolved hostname can just echo the label; still prefers an explicit HostKeyAlias over the dialed host; falls back to the dialed host when nothing was resolved; is rejoined when the split fragments resolve to nothing; leaves a single unresolvable path alone rather than inventing one; expands the __PROGRAMDATA__ token rather than passing it through as a path; does not rewrite a path that merely BEGINS with the token characters; keeps a drive-letter path with mixed separators intact; splits a spaced Windows home into the fragments the rejoin has to put back. src/main/ssh/config/ssh-known-hosts.test.ts: reads the algorithm from the blob rather than trusting the line; rejects %s; refuses a malformed blob: %s; uses the bare host on the default port; tries the bracketed form first and falls back to bare off-port; matches an exact line; matches case-insensitively; reports a changed key of the same type; reports unknown for %s; matches any host in a comma list; matches the bracketed form on its port; does not apply a bracketed entry to the default port; falls back to a bare line when off-port; never reports a change from the fallback pass; matches a bare literal on the default port; matches a bracketed literal on its port; reports a change for a literal; handles %s; still reports a change for a different key on a revoked host; skips a line carrying an unrecognised marker; does not report a change when only another key type is on file; matches within the same type; accepts CRLF terminators; skips blanks, comments and surrounding whitespace; accepts a hit regardless of which file holds it (%s); never brackets the alias, whatever the port; still brackets an ordinary host on a non-default port; matches a bare alias entry on a non-default port; still accepts the unmodified key field; accepts a hashed entry whose salt is a full SHA1 digest; still accepts the unmutated ssh-keygen -H vector; falls back to the bare host line on a non-default port, as ssh did; refuses a key type it has not seen for a host it knows, as ssh did. src/main/ssh/connection/ssh-connect-config.test.ts: returns undefined when no default keys exist; returns the first existing key file; probes regular and FIDO2 keys in stable default order; keeps a regular default ahead of a malformed FIDO2 default; uses target host/port/username; falls back to resolved config when target fields are empty; keeps imported endpoint fields as the fallback when ssh -G is unavailable; sets readyTimeout to CONNECT_TIMEOUT_MS; sets keepaliveInterval to 15s; enables keyboard-interactive auth so MFA challenges can be answered; uses agent auth when no explicit key and SSH_AUTH_SOCK is set; enables agent forwarding when OpenSSH config requests it and an agent is available; does not enable agent forwarding without a usable agent; resolves IdentityAgent SSH_AUTH_SOCK from the environment; uses the Windows OpenSSH agent pipe when no environment socket is available on Windows; wraps agent auth with IdentityFile filtering when IdentitiesOnly is enabled; applies a manual target IdentitiesOnly setting when ssh -G reports it off; includes unencrypted target.identityFile auth when an agent is available; defers encrypted target.identityFile auth when an agent is available; uses keyFile auth when target.identityFile is set and no agent is available; includes unencrypted resolved identityFile auth when an agent is available; defers a default-path resolved identityFile to the no-agent retry; offers the keyboard-interactive challenge when the agent defers no key; provides fallback key when no agent is available; can force private key inclusion for the post-agent fallback path; falls back to resolved proxyCommand; returns structured jump-host config for target.jumpHost; returns structured jump-host config for resolved proxyJump; returns undefined when no proxy is configured. src/main/ssh/connection/ssh-connection-auth-fallback.test.ts: resolves OpenSSH config using configHost when present; tries ssh-agent before reading an explicit private key; falls back to direct private key auth when agent auth fails; falls back to direct private key auth after too many agent authentication failures; answers bounded keyboard-interactive challenges such as Duo 2FA. src/main/ssh/connection/ssh-connection-store.test.ts: lists picker suppression aliases without consulting re-adoption tombstones; getTarget delegates to store; addTarget generates an id and persists; saves the CLI control switch as given, with no consent stamp; addTarget stamps source as manual by default; addTarget preserves an explicitly provided source; inserts a new config host stamped as ssh-config; asks the parser for all hosts — reconciliation happens in the store; updates an existing config-sourced target when the port changed; refreshes gssapiAuthentication on sync; never overwrites a manual target that owns the alias; does not rewrite an unchanged config-sourced target; returns empty array when nothing changed; tombstones a deleted config-sourced host so re-import skips it; suppresses a deleted manual target from config discovery; re-adding a deleted host reclaims its alias so sync stops suppressing it; editing a target reclaims its alias from the deleted set; records a tombstone when removing any user-facing target; re-adopts orphaned repos when the same host is re-added. src/main/ssh/connection/ssh-connection.test.ts: transitions to connected on successful connect; captures the negotiated SSH server key fingerprint; allows concurrent exec commands for ssh2 transport; removes startup listeners after ssh2 connect succeeds; scopes lifecycle events and pending handshake timers to one mock client; enables TCP_NODELAY on the new ssh2 client after a reconnect cycle; transitions through connecting → connected states; reports error state on connection failure; disconnect cleans up and sets state to disconnected; getTarget returns a copy of the target; getState returns a copy of the state; throws when connecting a disposed connection. src/main/ssh/orcad-migration-stuck-host-profiles.test.ts: never leaves a running terminal behind; is offered only when leaving saved state behind would clear the refusal. src/main/ssh/orcad-runtime-maintenance.test.ts: keeps a connect from retrying a build an explicit Update also failed; offers Force when Update was refused by policy; forwards anew after an activation, since the new slot may bind another port; passes a refusal through and leaves the server linked; forwards anew once a serving slot is restored, which may have bound another port. src/main/ssh/port-forwarding/ssh-port-forward.test.ts: reports the port the OS bound when an ssh2 forward asks for port 0; adds an ssh2 port forward and returns entry; adds a system SSH port forward when no ssh2 client is available; reports the port a system SSH forward resolved when asked for port 0; passes resolved OpenSSH config to system SSH port forwards; disposes a system SSH tunnel when removing the forward; lists forwards filtered by connectionId; removes a forward by id; awaits forward close when removing a forward for user actions; returns null when removing nonexistent forward; removes all forwards for a connection; dispose removes all forwards; stores label in the entry; preserves the existing id when updating a forward succeeds. src/main/ssh/ssh-channel-multiplexer.test.ts: sends a JSON-RPC request and resolves on response; assigns unique request IDs; sends notifications without expecting a response; dispatches incoming notifications to handler; sends one keepalive frame per cadence tick; drives keepalive sends AND dead-link detection from a single interval; resolves false immediately on a disposed mux; throws on request after dispose; reports isDisposed correctly; clears registered handlers on dispose; does not retain handlers registered after dispose. src/main/ssh/system-transport/system-ssh-forward-process.test.ts: spawns exactly the argv the orphan matcher is given, LogLevel=INFO included; runs port forwards unmultiplexed even when ssh_config disables muxing; preserves manual target port and identity options in the forwarded ssh command; does not spawn ssh when the requested local forward port is already in use; spawns ssh after the requested local forward port preflight succeeds; resolves startup after the grace period when ssh keeps running; sends SIGTERM then SIGKILL when the process does not exit; does not resolve stop until the process exits. src/main/ssh/system-transport/system-ssh-loopback-route.test.ts: dials the held forward ahead of the alias so OpenSSH keeps these over ssh_config; drops a manual target's own port and proxies, which would fight the route; never routes a forward, whose argv must stay stable for orphan recovery; %s → %s; opens the forward over the real proxy, then proves a login over the route; lands the forward on the sshd and known_hosts name the direct route resolved; holds the forward with server-alive probes; never lets the background forward or probe ask for credentials; starts no forward when admission refuses the route; stops the forward when sshd refuses logins over it; stops the whole forward process tree on Windows, once; reports a forward that already exited before anyone listened for it. src/main/startup/browser-process-user-agent-ordering.test.ts: explains admission refusal before a desktop launch exits; explains an orcad holding the profile instead of exiting silently; does not acquire profile admission for a duplicate launch; executes after the dev app name and before later preflight work; exits and releases admission after pending profile move recovery fails; keeps Linux background startup failures console-only and immediate; keeps Linux serve startup failures console-only and immediate. src/main/startup/quit/quit-path-durable-write-blocking.test.ts: store.flushAsync() issues no synchronous fs syscalls; store.flushAsync() also moves the active-view and github-cache sidecars off the thread; stats.flushAsync() issues no synchronous fs syscalls; Grok session cleanup issues no synchronous fs syscalls; stats.flushAsync() closes out live agents before it returns, not when its write lands; a stalled mount leaves the event loop live instead of parking it; store.flushAsync() is one idempotent final barrier; store.flushAsync() resolves instead of throwing when the profile mount rejects writes. src/main/stats/agent-session-transition-recorder.test.ts: counts a hook-only agent that never writes an agent-shaped terminal title; never opens a session from a replayed or disk-restored working status; still closes a live session when the terminating status arrives as a replay; counts one session per turn across repeated working/done cycles; closes an open session when its pane is torn down; bounds the mirror and closes what it evicts; stops the session when the main agent settles and only a watch loop holds the row working; keeps the session open while a subagent outlives the main agent, and closes it when the child settles; dates the stop by the evidence when a shell outlives the last subagent; opens a new session dated by the observation when the main agent resumes after monitoring; never dates a reopen before the close a fact-less OSC repaint caused; dates the first live row after a restored one by the evidence, not the restored clock; treats an unchanged answer as a snapshot, never a transition; opens only on a live working edge; closes only a session it opened. src/main/text-generation/commit-message-text-generation-model-discovery.test.ts: returns static catalog models without spawning for static agents; discovers dynamic models through the agent CLI; discovers dynamic models through the configured agent command override; falls back to static models when dynamic discovery returns no parseable models; falls back to the first discovered non-Pi model when its static default is unavailable. src/main/text-generation/pull-request-context.test.ts: does not fetch unrelated fork remotes before generating PR context; bounds exact remote-ref probe concurrency for a large remote list; reports no branch change because PR context preparation is read-only; keeps a remote-qualified base when the selected base includes the remote. src/main/text-generation/source-control-local-process.test.ts: asks the supervisor to stop and leaves its group to it; signals nothing once the supervisor has exited; stops an agent that ignores its stdin end through its supervisor. src/main/updater/linux-package/linux-package-update-recovery.test.ts: retains the downloaded package with the digest from the event metadata; requires a string downloaded path; accepts an uppercase extension; requires a non-empty string version; matches an absolute file URL by basename; accepts duplicate entries that agree on the digest; refuses an entry with a missing or non-string digest; refuses a digest that is not base64 at all; refuses a malformed percent-encoded URL; refuses a missing or non-array file list; tolerates a non-object event; replaces the retained artifact with a newer download; keeps the retained artifact when a new download has no usable digest; keeps the retained artifact when an unrelated download event arrives; clears the retained artifact; preserves the artifact across a recheck of the same version; clears the artifact when another version takes over; preserves the artifact for an unknown or empty version; is a no-op without a retained artifact; returns the built command and package file name for a verified package; reports missing without a retained artifact; propagates a package-manager discovery failure; rejects a well-formed digest that belongs to another file; never hashes a package that a malformed digest failed to arm; rejects a path outside the cache root; rejects the cache root itself; rejects a package in the updater directory itself; rejects a pending directory that is a symlink out of the cache; performs one hash pass for concurrent requests; revalidates once the in-flight pass settles; does not reuse an in-flight pass for a different artifact; starts a fresh proof when the same package is captured again; returns the verified package path; returns missing without a retained artifact. src/main/updater/releases/updater-prerelease-feed-readiness.test.ts: accepts a GitHub %s asset redirect as ready; reports a GitHub asset request error as unavailable; reports not-ready with a verified last-good tag when the newest assets are unavailable; does not classify a non-positive tag limit as feed unavailability; reproduces the v1.4.142 publishing incident as not-ready; reports transport failures as unavailable instead of not-ready; treats an explicit asset 404 as not-ready when another asset is unavailable; accepts absolute manifest asset URLs without rewriting them to release asset paths; returns not-ready with an older ready update as last-good while newest is publishing. src/main/updater/releases/updater-release-builds.test.ts: lists hourly builds from the dedicated repo, newest first; lists daily builds from the dedicated repo, newest first; separates stable from rc in the shared main repo; skips drafts and unparseable tags; keeps a composed release title and drops one that repeats the tag; hides builds that published no artifact for this platform; returns an empty list when no build has this platform artifact; keeps mac builds visible on macOS regardless of the Windows leg; resolves the platform installer download url; leaves the installer url null when the release published no installer; tolerates a release whose assets are missing or malformed; sends the local gh token as a bearer header when one is available; sends no authorization header when gh has no token; retries unauthenticated once when GitHub rejects the token; does not retry a 401 that was already unauthenticated; falls back to the per-IP bucket when the token is rate limited and tells the gh breaker; reports a secondary limit carrying retry-after without retrying or tripping the breaker; records a primary limit in the token scope %s; keeps %s authenticated when only %s is blocked; skips the token while its scope %s is blocked; surfaces a rate limit with its reset time and a sign-in hint when unauthenticated; omits the sign-in hint when the rate-limited request was authenticated; treats 429 as a rate limit; reports a 403 without rate-limit headers as a plain HTTP error; reports a missing hourly repo distinctly; is null when GitHub sent no reset; prefers retry-after over the primary reset epoch; falls back to the primary reset epoch when there is no retry-after; reads retry-after as seconds; reads retry-after as an HTTP date; pins an hourly tag at the hourly repo download path; pins a daily tag at the daily repo download path; pins a stable tag at the main repo download path; rejects a tag that is not a version. src/main/updater/updater.build-channel-selection.test.ts: uses the display label in the unsupported-platform %s pinned-build error; still pins %s from an unsigned Windows dev build; restores ordinary release checks after local build selection fails; opts into the RC channel when checkForUpdatesFromMenu is called with includePrerelease; pins the generic feed to a perf-tagged prerelease when requested; surfaces no-update feedback when no newer perf-tagged prerelease exists; keeps prerelease publishing-window misses on the generic retry path; leaves the feed URL alone for a normal user-initiated check. src/main/updater/updater.check-preflight.test.ts: ignores a stale update-not-available event after a new check starts preflight; deduplicates repeated manual checks while the immediate checking status is active. src/main/updater/updater.prerelease-fallback.test.ts: retries a prerelease check once against the previous feed tag when the manifest is missing; surfaces a promise-only prerelease fallback failure after the primary error event. src/main/updater/updater.publishing-window-feed.test.ts: repins the generic feed to the newest RC tag for a prerelease user; repins the generic feed to a newer stable tag for a prerelease user; pins a stable fallback when Atom has no newer prerelease; uses last-good concrete feed when a user-initiated check lands during publishing; can offer an older fully published update while the newest release is publishing; keeps background publishing-window fallback on the short retry cadence. src/main/updater/updater.quit-and-install.test.ts: surfaces an accepted retry before electron-updater emits download progress; defers quitAndInstall and runs its launcher within the launch PATH scope; runs pre-quit cleanup before local PTY cleanup during update install; keeps optional pre-quit cleanup fail-and-continue behavior by default; ignores duplicate quitAndInstall requests while the shared delay is pending; arms the forced-exit watchdog once the install commits; disarms the forced-exit watchdog when sync install error recovery keeps the app open. src/main/window/createMainWindow-markdown-editor-focus.test.ts: ignores traffic light sync IPC on non-macOS; still intercepts Cmd+Shift+B and Cmd+Alt+B when the markdown editor is focused; coerces non-boolean setMarkdownEditorFocused payloads to false; opens a table-aware context menu synchronously without a renderer query; builds the plain edit menu for the one right-click the renderer marks as plain text. src/main/window/createMainWindow-renderer-crash-recovery.test.ts: notifies the caller when the renderer process is gone; passes the renderer webContents id through crash recording and recovery callbacks; forwards expected renderer teardowns so the recorder can diagnose suppression; reloads the app shell after an unexpected renderer process loss; does not reload after renderer loss when recovery is disabled; stops auto-reloading after a rapid renderer crash loop trips the breaker. src/main/window/dashboard-popout-window.test.ts: updates the title after language initialization and unsubscribes when closed; lets its preload expose saved WSL hosts in serve-only, and passes nothing otherwise; creates a native-frame window with the shared preload and no webview surface; shows the window on ready-to-show; loads the prod file entry; loads the dev server URL when in dev; focuses the existing window instead of creating a second one; trusts only the live popout webContents; restores a minimized window when re-requested; restores valid persisted bounds; discards off-screen persisted bounds and falls back to defaults; persists bounds on resize after the debounce, guarding near-minimum sizes; keeps saving bounds after a vetoed quit and freezes them only on an allowed quit; closeDashboardPopout closes an open window; applies the persisted app zoom level on dom-ready; follows app zoom changes while open and unsubscribes on close; zoomDashboardPopoutIfFocused steps the popout zoom only while focused; handles the zoom-in chord via before-input-event and ignores other keys; handles mouse-wheel zoom requests outside before-input-event; respects zoom keybinding overrides for keyboard and mouse-wheel paths. src/main/window/dropped-path-preparation.test.ts: opens and stores each unique source once as drop size grows tenfold; passes ordinary paths without copying for %s; preserves ordinary successes and copied successes when another file fails; serializes copy requests from different owners; reports all lost paths without a shared reason when their failures differ. src/main/window/main-window-updater-deferred-setup.test.ts: releases all 64 completed fallback handles without repeating initialization; keeps setup after ready-to-show and releases its fallback afterward; preserves the 15-second fallback when first paint never arrives; preserves destroyed-window guard reads before and at the fallback deadline; keeps the exact %s error and releases its completed fallback; keeps the exact fallback %s error without retrying initialization; preserves pending manual setup after failed ready registration; keeps setup options, live store reads, nudge writes and required quit flush. src/main/worker-threads/worker-thread-request-queue.test.ts: posts one request at a time and in the order it was dispatched; fires a posted call at its own deadline; stops respawning and fails the rest of the queue once deaths hit the cap; clears the death count when a fresh burst starts from full idle. src/main/workspace-hooks/hooks-setup-runner-script.test.ts: builds Windows runners for newline-heavy scripts without line-array splitting; builds POSIX runners without regex-wide CRLF normalization; writes POSIX setup runners for shebang-declared scripts on native Windows paths; keeps batch setup scripts on cmd.exe when the terminal is Git Bash; hands a Git Bash pane a launch command MSYS cannot rewrite for a cmd runner; replays interpreter flags declared on the shebang line; preserves cmd.exe setup runner semantics for configured cmd users; keeps POSIX runner behavior on POSIX platforms; waits when either local or project policy requires completed setup; marks setup-runner terminals for the always-on credential guard; maps git-bash to POSIX setup launch metadata on Windows; falls back to the cmd runner when the git-bash setting has no installed Git Bash; keeps the cmd runner for a non-Git bash such as Cygwin; keeps the cmd runner for a bare bash whose flavor cannot be resolved; uses the POSIX runner for a bare bash that resolves to Git Bash; uses the POSIX runner for an extension-less Git Bash path; preserves the existing cmd runner for PowerShell terminals; preserves cmd setup compatibility when a Windows-host project has a WSL shell setting; classifies an installed explicit Git Bash executable as a POSIX runner; falls back to the cmd runner when the explicit Git Bash path no longer exists. src/main/workspaces/catalog/repo-git-remote-identity-enrichment.test.ts: schedules remote identity enrichment without blocking the caller; probes a drive project whose runtime is WSL %s; sweeps a local repo: %s; probes an SSH row that carries only executionHostId on its own host; keeps same-path rows on two different SSH hosts from sharing one backoff; coalesces concurrent probes for the same repo location; caches no-identity probes briefly so list calls do not retry every time; does not rewrite the no-remote marker on a later retry; resolves a settled no-remote repo once it gains a remote; does not re-probe a resolved identity before the refresh window elapses; re-probes a resolved identity after the refresh window elapses; bounds re-probes per sweep so a large repo list drains gradually; forgets a removed repo location so its deadline cannot outlive the repo; keeps a surviving repo backoff when a sibling repo is removed; probes again when the same location is re-added after its probe was retired; coalesces concurrent sweeps instead of running one pass per list call. src/main/workspaces/creation/preparation/worktree-create-preparation.test.ts: preserves create priority through claim probes on %s; starts the checkout only once the async workspace root resolves; still deduplicates when both callers await the same pending root lookup; namespaces native Windows preparation directories for long paths; deduplicates preparation for the same repo, base, runtime, and workspace root; does not spend a divergence walk when the base matches exactly; claims when the two sides spell the same ref differently; evicts a repo's own stale preparation before another repo's; still discards armed checkouts when the project runtime needs repair; routes preparation and finalization through the selected WSL runtime; retries stale cleanup after a transient listing failure; does not discard a detached worktree with caller-controlled preparation metadata; does not re-arm after an isolated create; re-arms a preparation once creates arrive in a burst; holds the re-arm checkout until the create runs the deferred thunk; starts one explicit prefetch after the create completes; does not re-arm when finalization failed; reports a pending create while a stale-cleanup scan is running. src/main/workspaces/removal/local-worktree-removal-recovery.test.ts: recovers Git for Windows partial filesystem deletion failures; finishes a clean non-force deletion after Git leaves a Windows directory behind; recovers localized Windows failures after Git already removed the registration; does not recover unrelated localized failures while Git still owns the row; finishes a macOS deletion Git deregistered before failing with Directory not empty; leaves an orphaned checkout to the orphan cleanup instead of deleting it; prunes and strictly verifies on the selected WSL host without deleting files or branches. src/main/workspaces/removal/worktree-removal-safety.test.ts: rejects deleting a worktree that contains another registered worktree; rejects POSIX home directories even when host homedir has a different path shape; refuses a home the host reported even when no path rule recognises it; refuses the Windows host's home directory from a POSIX client; refuses the Windows host's home directory even when the host reported nothing. src/main/wsl/discovery/distro-catalog.test.ts: retries asynchronous discovery after a transient wsl.exe failure; retries synchronous discovery after a transient wsl.exe failure; retries asynchronous discovery after wsl.exe reports no distros yet; retries synchronous discovery after wsl.exe reports no distros yet; bounds the wsl.exe spawn rate while no distros are installed; bounds the asynchronous spawn rate while no distros are installed; still caches a non-empty distro list for the process lifetime; re-probes a docker-desktop-only machine once a user distro can appear; backs off while the list keeps coming back empty; backs off after repeated distro-list failures; caps the empty-list backoff at five minutes; keeps reporting an empty result after it goes stale; keeps reporting a known distro list indefinitely; bounds cached homes while retaining the most recently used distros; still lists distros, and an excluded distro home probe fails like any other probe; gives the async home probe time for a cold boot instead of killing it at 5 s; retries availability after a probe timeout instead of latching false; bounds the probe rate while the failure window is open; caches a successful probe for the process lifetime; names an explicit spawn directory, so no deleted cwd can read as ENOENT; shares one wsl.exe spawn between concurrent probes; keeps reporting the last observed answer after it goes stale; holds a definitive failure far longer than a timeout when %s; backs off after repeated probe failures; re-probes a seeded retryable failure once its window lapses; re-probes availability once a distro list succeeds after %s; does not drop an availability failure for an empty distro list; keeps reporting unavailable off Windows without probing. src/main/wsl/discovery/wsl-distro-holder.test.ts: runs a busybox-safe loop through --exec and names the distro; spawns a detached-style holder, records it, and waits for the distro to run; a holder exiting with code 1 notes its distro may have stopped (E9); rejects a pid whose command line is not our holder for that distro; cleans up stale and corrupt records; keeps a holder and its record when release cannot verify it; reports unverifiable when the command line is unreadable even with a matching pid; release kills a verified holder and removes its record; release never kills a pid that is no longer our holder; reports a spawned holder exit so liveness can re-check. src/main/wsl/filesystem/wsl-guest-drive-mountinfo.test.ts: skips 9p shares that are not drvfs, and malformed lines. src/main/wsl/orcad/wsl-orcad-deploy-app-update.test.ts: plans the update on the census the running orcad reports, not an unknown one; forwards force to the update planner; does not serve the incumbent past a refusal that is not a wait; does not re-activate the build this host rolled back from; labels the launch with the build the adoption found serving, not the stale record; defers while another run holds the host fence; lets an explicit force deploy past both skips; names a newer host refusal apart from a rolled-back one, so only Force is offered; updates an older host as before, without asking to adopt it; never asks the incumbent when it already runs this build. src/renderer/src/components/agent/agent-launch-picker-acquisition.test.ts: keeps a completed listing usable when the host rejects the wait parameter. src/renderer/src/components/feature-tips/feature-tip-telemetry.test.ts: delivers to chat-default users in all targeting while holding out not-switched users. src/renderer/src/components/native-chat/composer/pickers/use-native-chat-side-chat-command.test.ts: completes the row on a pick, even when the command is already typed; ignores a pick of the row while side chats are unavailable. src/renderer/src/components/native-chat/media/images/native-chat-image-copy.test.ts: explains a too-large refusal that arrives through the IPC error prefix. src/renderer/src/components/native-chat/messages/actions/native-chat-fork-launch.test.ts: keeps the question with the caller and says once that it never went; tells the user to check the side chat when delivery is unconfirmed; reports a delivered question as sent, silently. src/renderer/src/components/native-chat/messages/actions/native-chat-side-chat-fork-point.test.ts: decides each of %i tool-only turns once per layout, and nothing without a listener. src/renderer/src/components/native-chat/messages/actions/use-native-chat-fork.test.tsx: places a side chat whose question was %s: %o. src/renderer/src/components/terminal-pane/provider-notices/agent-presence/agent-terminal-presence.test.ts: calls back once for %s; stops watching once unsubscribed. src/renderer/src/components/terminal-pane/pty-connection/emulation/pty-connection-mode2031-raw-subscription.spec.ts: records a mode 2031 subscribe from the raw chunk boundary without answering. src/renderer/src/components/terminal-pane/pty-connection/input/pty-connection-deferred-input-forwarding.spec.ts: routes terminal input through deferPtyInput so the host can withhold it; forwards terminal input directly when the host supplies no deferPtyInput. src/renderer/src/components/terminal-pane/pty-connection/notifications/pty-connection-terminal-bell-attention.spec.ts: forwards onBell to pane attention and notification policy; hands the bell to the attention module with its pane even when pane attention is off. src/renderer/src/components/terminal-pane/pty-connection/output-restore/pty-connection-dropped-output-snapshot-healing.spec.ts: repaints from the main-owned snapshot when main drops pending output at the cap. src/renderer/src/components/terminal-pane/pty-connection/output/pty-connection-fresh-native-follow-reset.spec.ts: retries a fresh-spawn native follow reset when renderer dimensions return. src/renderer/src/components/terminal-pane/pty-connection/spawn/unbound-pane-spawn-recovery.test.ts: requests a tab remount for an unbound spawn; leaves recovery to a remote transport that armed its own retry; requests a remount when the transport reports no recovery of its own; settles the spawn as failed before requesting a remount; prefers the restored leaf id when reporting the anomaly; falls back to the pane leaf id when no restored leaf exists. src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-foreground-agent.test.ts: remove only literal registration import/call pairs import { registerPtyConnectionDeferredInputForwardingTests } from '../input/pty-connection-deferred-input-forwarding.spec', registerPtyConnectionDeferredInputForwardingTests(); no additional authored credit. src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-hidden-replay.test.ts: structural support changed only; no new authored removal credit. src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-input-and-repaint.test.ts: remove only literal registration import/call pairs import { registerPtyConnectionFreshNativeFollowResetTests } from '../output/pty-connection-fresh-native-follow-reset.spec', registerPtyConnectionFreshNativeFollowResetTests(); no additional authored credit. src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-spawn-and-startup.test.ts: remove only literal registration import/call pairs import { registerPtyConnectionTerminalBellAttentionTests } from '../notifications/pty-connection-terminal-bell-attention.spec', registerPtyConnectionTerminalBellAttentionTests(); no additional authored credit. src/renderer/src/components/terminal-pane/pty-transport/delivery/pty-dispatcher-delivery-interest.test.ts: registers interest on the first sidecar and releases on the last unsubscribe; releases sidecar interest only once for repeated unsubscribes; does not let an eager pre-mount buffer defeat hidden delivery gating; keeps eager buffers outside delivery interest when the PTY exits before mount; lets a sidecar exclusively own interest while an eager buffer overlaps. src/renderer/src/components/terminal-pane/pty-transport/delivery/pty-dispatcher-push-reattach.test.ts: blackholed delivery reproduces the wedge: no handler dispatch, no ACK ever. src/renderer/src/components/terminal-pane/pty-transport/ipc/ipc-pty-opencode-model.test.ts: sends preferences to the execution host and attaches only to its returned PTY; preserves one operation ID across a lost reply and reconnect; preserves ordinary raw startup and captured-session attachment. src/renderer/src/components/terminal-pane/pty-transport/output/pty-side-effect-pending-census.test.ts: sums registered gauges and drops disposed ones; ignores a repeated dispose instead of dropping another gauge; stops tracking new gauges past the cap so a missed dispose cannot grow it; tracks a live output processor queue through enqueue, drain, and dispose; excludes cleared consumed entries from retention after a bounded drain. src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-prebind-discard-notice.test.ts: says nothing once the pane is closed. src/renderer/src/components/terminal-pane/render-health/terminal-webgl-atlas-recovery.test.ts: refreshes after each scheduled atlas reset; falls back to a timeout when animation frames are unavailable; continues recovery when a manager throws after scheduling. src/renderer/src/components/terminal-pane/snapshots/pty-buffer-serializer.test.ts: tracks a remounted quiet title source even when the stale owner unregisters later; preserves the renderer-parsed output sequence in the response; preserves a pending escape tail for snapshot replay; routes a host's Reset Terminal request to the pane that owns the PTY. src/renderer/src/components/terminal-pane/snapshots/saved-terminal-output-replay.test.ts: marks panes with restored scrollback for fresh-shell viewport blanking. src/renderer/src/components/terminal-pane/viewport/split-right-white-screen.test.ts: recovers a 0x0 spawn whose container is briefly unmeasurable then settles. src/renderer/src/components/terminal-pane/viewport/visibility/stale-document-visibility.test.ts: removes global listeners when the last pane unregisters. src/renderer/src/components/ui/security-surface-theme-contrast.test.tsx: keeps trust text readable on a host-owned dialog surface; applies the hostile surface to ordinary dialogs, so the case above is exercised. src/renderer/src/components/use-terminal-parking-pass.test.tsx: does not queue a render when a pass produces the same parking sets; keeps parking transitions working under StrictMode. src/renderer/src/hooks/client-workspace-view-report.test.ts: reports an away user’s screen with no presence, and renews as soon as input returns; retries a failed send, empty or not, no sooner than a refresh. src/renderer/src/hooks/ipc-events/terminal-request-ipc-bridge-placement.test.ts: is new work in a workspace with no live terminal: its first bind moves it up in Recent. src/renderer/src/hooks/use-bundled-skill-install.test.tsx: ignores a repeated click while the install runs; retries the last requested skills on this host. src/renderer/src/hooks/useGitHubSlugMetadata.assignee-reuse.test.tsx: adds seeds to a successful empty list but not to loading or failed results; retains seed projection after expiry and fetches fresh metadata for a new reader. src/renderer/src/hooks/useSettingsNavigationMetadata.language-switch.test.tsx: retranslates the settings nav labels when the UI language changes live. src/renderer/src/lib/codex-stale-pane-account-identity.test.ts: keeps the prompt when the roster read fails and every label collapses. src/renderer/src/lib/image-blob-png.test.ts: downscales a pasted image above the pixel cap; releases its temporary URL on a decode error; releases its temporary URL when PNG encoding fails or exceeds the byte cap. src/renderer/src/lib/react-commit-cascade-store-write-samples.test.ts: names callers above the frame the write came from; restores Error.stackTraceLimit when the capture throws; stops collecting after the sample cap when captureStackTrace is unavailable; caps reported changed keys from a single wide write. src/renderer/src/lib/work-items/linear/linear-issue-context-snapshot.test.ts: applies the rendered text hard cap last and reserves the final overflow marker. src/renderer/src/runtime/rpc/targets/use-worktree-runtime-target.test.ts: keeps the runtime target identity stable across unrelated store writes. src/renderer/src/runtime/source-control/git/runtime-git-branch-client.test.ts: keeps creation available on old hosts and explains the missing operation; preserves landed warnings and direct collision recovery data; propagates remote discovery failure instead of a false empty page. src/renderer/src/runtime/sync-runtime-graph/sync-runtime-graph-key-reuse.test.ts: reuses unchanged worktree tab projections when one worktree title changes; bumps only the changed worktree version when a sibling stays byte-identical. src/renderer/src/store/repos/repos-catalog-request-coordination.test.ts: shares safe-auto fork sync cooldown attempts across store instances. src/renderer/src/store/slices/tabs/hydrated-workspace-reconciliation-batch.test.ts: collapses a 193-workspace hydration to one store write; leaves the store byte-identical to the per-workspace path; re-reconciling the batched result is a no-op, as it is for the per-workspace path. src/renderer/src/store/slices/tabs/hydrated-workspace-reconciliation-fixture.ts: dedicated zero-case helper retirement; no authored test credit. src/renderer/src/store/slices/worktrees/metadata/update-worktrees-meta.test.ts: persists each batch metadata update to its workspace owner. src/renderer/src/store/slices/worktrees/metadata/worktree-meta-rename.test.ts: dismisses a rename error on its capable remote owner; clears pending first-agent rename when the title is updated. src/renderer/src/store/terminals/store-sleep-shutdown-order.test.ts: orders capture, guards, handlers, main's stop, commit, cleanup, and exit settlement. src/renderer/src/store/terminals/terminal-first-tab-publication.test.ts: publishes the first terminal and root tab group atomically; publishes the first terminal and root tab group atomically, then marks it starting in main. src/renderer/src/store/terminals/workspace-terminal-reconnect-hydration.test.ts: restores daemon terminal wake hints before declaring the session ready. src/renderer/src/web/web-runtime-connection-auth-framing.test.ts: keeps legacy framing for sockets that cannot absorb a refused attempt. src/shared/terminal/appearance/terminal-font-weight-windows-regular.test.ts: moves the old Windows default to Regular once and queues the notice; keeps custom weights and fresh profiles without a notice; leaves macOS and Linux profiles untouched. tests/e2e/claude-agent-row-view.unit.test.ts: updates a mounted compact workflow description through the subscribed production parent; bounds upstream active-SSH visits for $population records, fallback=$fallback; derives %s destinations without requiring type or a workflow child ID; terminates a $path parent $status with no third request or intermediate toast; does not recover a first child %s. tests/e2e/cross-version-wire/agent-permission-legacy-spawn.unit.test.ts: keeps exact legacy ordering when structured creation cannot express afterTabId; can create a terminal without selecting the target worktree; preserves the legacy fresh-agent path when host authority is unavailable. tests/e2e/paired-instant-browser-tab.spec.ts: shows a paired browser tab before its create RPC resolves, then keeps it as one tab; adopts a paired browser tab without rebuilding its chrome. tests/e2e/session-tabs-decorative-title-fanout.unit.test.ts: bounds host publication and renderer application across remote worktrees; globally caps decorative freshness while every working status stays fresh. tests/e2e/session-tabs/mobile-session-active-performance.unit.test.ts: bounds saved-order publication selection at N and 10N without rematerializing navigation; bounds admitted placement reconciliation during supported reads at N and 10N; bounds selection input visits at 50 and 500 rows, including unavailable-target projection; keeps publication scoped to a changed worktree and shared by listeners at N and 10N; bounds preserved-chat merge and runtime-owned hydration at N and 10N (legacy=%s); bounds actual native retirement and shared delivery work at N and 10N. tests/e2e/session-tabs/mobile-session-active-publication.unit.test.ts: does not substitute a preserved row for an unavailable explicit legacy selection; resolves main-layout group selection %s (omitted group=%s). tests/e2e/structured-agent-session-launch-admission.unit.test.ts: opens a direct caller's terminal with the caller's own CLI arguments on a decline; opens the chat on the server that admitted it; opens only the terminal when this machine declines, without a notice; opens a new agent tab's terminal with the typed prompt when the caller names none; opens the chat, not a terminal, when this machine cannot resolve a new workspace yet; keeps its terminal fallback, as before. tests/e2e/structured-attention-prompt-relay.unit.test.ts: settles once per relayed prompt, not on every later status update. tests/e2e/structured-attention-read-retirement.unit.test.ts: Mark read on a chat never opened withdraws the prompt alert it surfaced; Mark read on a chat hidden since an earlier view covers the newer prompt it surfaced; reads a newly accepted visible prompt without sending an RPC for text-only updates. tests/e2e/structured-chat-tab-lifetime-cost.unit.test.ts: counts real raw stores, unrelated window publications and captured cleanup at %i chats; counts complete supplied-member sweep and separate final retirement at S=%i N=%i. tests/e2e/structured-session-workspace-authority.unit.test.ts: bounds actual target reveal across %i repo roots and metadata aliases. tests/e2e/terminal-agent-pane-launch.unit.test.ts: starts the host's planned command once, at the pane's size, and says it started; plans a pane whose prompt the window submits without the startup update prompt; preserves caller shell-ready delivery only for a connection-scoped adapter ($connectionId); starts no shell on plan failure (original failure: $returnLaunchFailure); cannot confirm a start whose provider answer was lost. tests/e2e/terminal-parked-memory.spec.ts: releases renderer terminal memory when hidden tabs park. tests/e2e/terminal-shortcuts.spec.ts: @headful Codex-like background output stays visible without disabling WebGL in auto mode; Shift with Russian layout text reaches the PTY as Cyrillic under kitty keyboard reporting. tests/e2e/terminal-split-activation-latency.spec.ts: records attributed CWD, spawn, bind, fixture-ready, input, and echo phases. tests/e2e/terminal-split-activation-latency-phases.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-split-activation-latency-main-probe.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-split-activation-latency-artifact.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-split-activation-latency-report.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-tab-switch-visual-restore.spec.ts: restores skipped hidden Grok output on light tab resume. tests/e2e/web-runtime-session-terminal-host-authority.unit.test.ts: delivers independent %s effort through the real prepared web callback; keeps same-ID local and runtime worktrees on the selected runtime owner; creates paired web agents through host authority so activation is mirrored; delivers generated continuation context after host-authoritative creation; seeds the chat composer for a draft that rode in on the launch command. Current changed source bytes, wrapper registrations and reduced selectors have no fresh execution or red/green credit. Unrelated surviving receipt objects keep their original narrow scope; current literal graph reachability does not establish historical import closure or equivalent replacement coverage. Current conservative command/import graph reaches src/main/runtime/construction/orca-runtime.test.ts, src/main/runtime/orca-runtime-tests/targeting-and-resilience.spec.ts, src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-04.spec.ts, src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-03.spec.ts, src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-02.spec.ts, src/main/runtime/orca-runtime-tests/gitlab-and-pr-bases-part-02.spec.ts, src/main/runtime/orca-runtime-tests/gitlab-and-pr-bases.spec.ts, src/main/runtime/orca-runtime-tests/worktree-setup-and-startup-part-03.spec.ts, src/main/runtime/orca-runtime-tests/worktree-setup-and-startup.spec.ts, src/main/runtime/orca-runtime-tests/terminal-sleep-and-teardown-part-02.spec.ts, src/main/runtime/orca-runtime-tests/mobile-summaries-part-03.spec.ts, src/main/runtime/orca-runtime-tests/mobile-summaries-part-02.spec.ts, src/main/runtime/orca-runtime-tests/mobile-summaries.spec.ts, src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-03.spec.ts, src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-02.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-13.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-12.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-11.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-10.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-08.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-07.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-05.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-04.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-02.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-05.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-04.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-03.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-02.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status.spec.ts, src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-07.spec.ts, src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-04.spec.ts, src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-03.spec.ts, src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-02.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-12.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-11.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-10.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-09.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-08.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-07.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-06.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-03.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness.spec.ts, src/main/runtime/orca-runtime-tests/headless-snapshots.spec.ts, src/main/runtime/orca-runtime-tests/terminal-side-effect-facts-part-02.spec.ts, src/main/runtime/orca-runtime-tests/local-worktree-creation-part-02.spec.ts, src/main/runtime/orca-runtime-tests/local-worktree-creation.spec.ts, src/main/persistence/worktree-sort-order-write.test.ts, src/main/runtime/integrations/skill/orca-runtime-skill-recovery.test.ts, src/main/runtime/orchestration/mailbox/routing/orchestration-mailbox-detached-routing.spec.ts, src/main/runtime/orchestration/mailbox/terminal-pointer/orchestration-mailbox-delivery.test.ts, src/main/runtime/orchestration/mailbox/terminal-pointer/orchestration-mailbox-pointer-cli-command.spec.ts, src/main/runtime/rpc/methods/terminal/terminal-multiplex-link-estimate.test.ts, src/main/runtime/rpc/methods/terminal/terminal-snapshot-scrollback-rows.test.ts, src/main/runtime/rpc/server/pairing/network/windows-default-route-interfaces.test.ts, src/main/runtime/rpc/server/pairing/pairing-offer-owner-cli.test.ts, src/main/runtime/rpc/terminal-multiplex-end-verdict.test.ts, src/main/runtime/session-graph/headless-runtime-graph.spec.ts, src/main/runtime/session-graph/runtime-graph-standing-contracts.test.ts, src/main/runtime/session-tabs/structured-conversation-tab-title.test.ts, src/main/runtime/session-tabs/terminal-close-observed-exit.test.ts, src/main/runtime/terminal/identity/terminal-identity-probe.test.ts, src/main/runtime/terminal/layout/paired-desktop-server-writes-layout-owner.test.ts, src/main/runtime/terminal/lifecycle/pty-lifecycle-generation-retention.test.ts, src/main/runtime/terminal/lifecycle/run-terminal-user-open.test.ts, src/main/runtime/terminal/surface/terminal-surface-replies-window.test.ts, src/main/runtime/terminal/surface/workspace-layout-projection.test.ts, src/main/runtime/terminal/view/terminal-subscribe-exit-waiter-leak.test.ts, src/main/runtime/terminal/wait/orca-runtime-exact-worker-provider-session.test.ts, src/main/runtime/terminal/wait/prime-agent-screen-readiness-transcripts.test.ts, src/main/runtime/workspaces/startup/runtime-worktree-agent-startup.test.ts, src/main/runtime/workspaces/startup/runtime-worktree-startup-readiness.test.ts, src/main/skills/skill-install-transaction.test.ts, src/main/speech/model-manager-download-resume.test.ts, src/main/ssh/config/ssh-config-parser.test.ts, src/main/ssh/config/ssh-known-hosts-source.test.ts, src/main/ssh/config/ssh-known-hosts.test.ts, src/main/ssh/connection/ssh-connect-config.test.ts, src/main/ssh/connection/ssh-connection-auth-fallback.test.ts, src/main/ssh/connection/ssh-connection-store.test.ts, src/main/ssh/connection/ssh-connection.test.ts, src/main/ssh/orcad-migration-stuck-host-profiles.test.ts, src/main/ssh/orcad-runtime-maintenance.test.ts, src/main/ssh/port-forwarding/ssh-port-forward.test.ts, src/main/ssh/ssh-channel-multiplexer.test.ts, src/main/ssh/system-transport/system-ssh-forward-process.test.ts, src/main/ssh/system-transport/system-ssh-loopback-route.test.ts, src/main/startup/browser-process-user-agent-ordering.test.ts, src/main/startup/quit/quit-path-durable-write-blocking.test.ts, src/main/stats/agent-session-transition-recorder.test.ts, src/main/text-generation/commit-message-text-generation-model-discovery.test.ts, src/main/text-generation/pull-request-context.test.ts, src/main/text-generation/source-control-local-process.test.ts, src/main/updater/linux-package/linux-package-update-recovery.test.ts, src/main/updater/releases/updater-prerelease-feed-readiness.test.ts, src/main/updater/releases/updater-release-builds.test.ts, src/main/updater/updater.build-channel-selection.test.ts, src/main/updater/updater.check-preflight.test.ts, src/main/updater/updater.prerelease-fallback.test.ts, src/main/updater/updater.publishing-window-feed.test.ts, src/main/updater/updater.quit-and-install.test.ts, src/main/window/createMainWindow-markdown-editor-focus.test.ts, src/main/window/createMainWindow-renderer-crash-recovery.test.ts, src/main/window/dashboard-popout-window.test.ts, src/main/window/dropped-path-preparation.test.ts, src/main/window/main-window-updater-deferred-setup.test.ts, src/main/worker-threads/worker-thread-request-queue.test.ts, src/main/workspace-hooks/hooks-setup-runner-script.test.ts, src/main/workspaces/catalog/repo-git-remote-identity-enrichment.test.ts, src/main/workspaces/creation/preparation/worktree-create-preparation.test.ts, src/main/workspaces/removal/local-worktree-removal-recovery.test.ts, src/main/workspaces/removal/worktree-removal-safety.test.ts, src/main/wsl/discovery/distro-catalog.test.ts, src/main/wsl/discovery/wsl-distro-holder.test.ts, src/main/wsl/filesystem/wsl-guest-drive-mountinfo.test.ts, src/main/wsl/orcad/wsl-orcad-deploy-app-update.test.ts, src/renderer/src/components/agent/agent-launch-picker-acquisition.test.ts, src/renderer/src/components/feature-tips/feature-tip-telemetry.test.ts, src/renderer/src/components/native-chat/composer/pickers/use-native-chat-side-chat-command.test.ts, src/renderer/src/components/native-chat/media/images/native-chat-image-copy.test.ts, src/renderer/src/components/native-chat/messages/actions/native-chat-fork-launch.test.ts, src/renderer/src/components/native-chat/messages/actions/native-chat-side-chat-fork-point.test.ts, src/renderer/src/components/native-chat/messages/actions/use-native-chat-fork.test.tsx, src/renderer/src/components/terminal-pane/provider-notices/agent-presence/agent-terminal-presence.test.ts, src/renderer/src/components/terminal-pane/pty-connection/emulation/pty-connection-mode2031-raw-subscription.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/input/pty-connection-deferred-input-forwarding.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/notifications/pty-connection-terminal-bell-attention.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/output-restore/pty-connection-dropped-output-snapshot-healing.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/output/pty-connection-fresh-native-follow-reset.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/spawn/unbound-pane-spawn-recovery.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-foreground-agent.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-hidden-replay.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-input-and-repaint.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-spawn-and-startup.test.ts, src/renderer/src/components/terminal-pane/pty-transport/delivery/pty-dispatcher-delivery-interest.test.ts, src/renderer/src/components/terminal-pane/pty-transport/delivery/pty-dispatcher-push-reattach.test.ts, src/renderer/src/components/terminal-pane/pty-transport/ipc/ipc-pty-opencode-model.test.ts, src/renderer/src/components/terminal-pane/pty-transport/output/pty-side-effect-pending-census.test.ts, src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-prebind-discard-notice.test.ts, src/renderer/src/components/terminal-pane/render-health/terminal-webgl-atlas-recovery.test.ts, src/renderer/src/components/terminal-pane/snapshots/pty-buffer-serializer.test.ts, src/renderer/src/components/terminal-pane/snapshots/saved-terminal-output-replay.test.ts, src/renderer/src/components/terminal-pane/viewport/split-right-white-screen.test.ts, src/renderer/src/components/terminal-pane/viewport/visibility/stale-document-visibility.test.ts, src/renderer/src/components/ui/security-surface-theme-contrast.test.tsx, src/renderer/src/components/use-terminal-parking-pass.test.tsx, src/renderer/src/hooks/client-workspace-view-report.test.ts, src/renderer/src/hooks/ipc-events/terminal-request-ipc-bridge-placement.test.ts, src/renderer/src/hooks/use-bundled-skill-install.test.tsx, src/renderer/src/hooks/useGitHubSlugMetadata.assignee-reuse.test.tsx, src/renderer/src/hooks/useSettingsNavigationMetadata.language-switch.test.tsx, src/renderer/src/lib/codex-stale-pane-account-identity.test.ts, src/renderer/src/lib/image-blob-png.test.ts, src/renderer/src/lib/react-commit-cascade-store-write-samples.test.ts, src/renderer/src/lib/work-items/linear/linear-issue-context-snapshot.test.ts, src/renderer/src/runtime/rpc/targets/use-worktree-runtime-target.test.ts, src/renderer/src/runtime/source-control/git/runtime-git-branch-client.test.ts, src/renderer/src/runtime/sync-runtime-graph/sync-runtime-graph-key-reuse.test.ts, src/renderer/src/store/repos/repos-catalog-request-coordination.test.ts, src/renderer/src/store/slices/tabs/hydrated-workspace-reconciliation-batch.test.ts, src/renderer/src/store/slices/tabs/hydrated-workspace-reconciliation-fixture.ts, src/renderer/src/store/slices/worktrees/metadata/update-worktrees-meta.test.ts, src/renderer/src/store/slices/worktrees/metadata/worktree-meta-rename.test.ts, src/renderer/src/store/terminals/store-sleep-shutdown-order.test.ts, src/renderer/src/store/terminals/terminal-first-tab-publication.test.ts, src/renderer/src/store/terminals/workspace-terminal-reconnect-hydration.test.ts, src/renderer/src/web/web-runtime-connection-auth-framing.test.ts, src/shared/terminal/appearance/terminal-font-weight-windows-regular.test.ts, tests/e2e/claude-agent-row-view.unit.test.ts, tests/e2e/cross-version-wire/agent-permission-legacy-spawn.unit.test.ts, tests/e2e/paired-instant-browser-tab.spec.ts, tests/e2e/session-tabs-decorative-title-fanout.unit.test.ts, tests/e2e/session-tabs/mobile-session-active-performance.unit.test.ts, tests/e2e/session-tabs/mobile-session-active-publication.unit.test.ts, tests/e2e/structured-agent-session-launch-admission.unit.test.ts, tests/e2e/structured-attention-prompt-relay.unit.test.ts, tests/e2e/structured-attention-read-retirement.unit.test.ts, tests/e2e/structured-chat-tab-lifetime-cost.unit.test.ts, tests/e2e/structured-session-workspace-authority.unit.test.ts, tests/e2e/terminal-agent-pane-launch.unit.test.ts, tests/e2e/terminal-parked-memory.spec.ts, tests/e2e/terminal-shortcuts.spec.ts, tests/e2e/terminal-split-activation-latency.spec.ts, tests/e2e/terminal-split-activation-latency-phases.ts, tests/e2e/terminal-split-activation-latency-main-probe.ts, tests/e2e/terminal-split-activation-latency-artifact.ts, tests/e2e/terminal-split-activation-latency-report.ts, tests/e2e/terminal-tab-switch-visual-restore.spec.ts, tests/e2e/web-runtime-session-terminal-host-authority.unit.test.ts. Unresolved local/computed edges remain explicit debt.", diff --git a/config/scripts/ci/sharding/ci-shard-timings.json b/config/scripts/ci/sharding/ci-shard-timings.json index 78bbe728ff8..9720bdb395b 100644 --- a/config/scripts/ci/sharding/ci-shard-timings.json +++ b/config/scripts/ci/sharding/ci-shard-timings.json @@ -235,7 +235,6 @@ "config/scripts/typecheck/project-execution/run-typecheck-projects-in-parallel.test.mjs": 182, "config/scripts/tool-invocation/script-entry-detection.test.mjs": 107, "config/scripts/test-fixtures/script-copy-closure/script-module-dependencies.test.mjs": 62, - "config/scripts/orcad-distribution/server-bundle/session-scanner-service-smoke.test.mjs": 562, "config/scripts/ci/workflow-contracts/setup-pnpm-action.test.mjs": 421, "config/scripts/electron-installation/shared-dist-cache/shared-electron-dist-cache.test.ts": 140, "config/scripts/native-build/node-addons/shared-native-build.test.mjs": 127, @@ -2186,12 +2185,9 @@ "src/main/git/remotes/worktree-push-target-setup.test.ts": 491, "src/main/ipc/worktree-symlink-reconciliation.real.test.ts": 410, "src/main/ipc/worktree-symlinks.test.ts": 152, - "src/main/ipc/worktrees-delete-pty-teardown.test.ts": 1934, - "src/main/ipc/worktrees-issue-command-overrides.test.ts": 1594, "src/main/ipc/worktrees-orphan-directory-cleanup.test.ts": 1486, "src/main/ipc/worktrees-removal-recovery.test.ts": 1971, "src/main/ipc/worktrees-remove-preflight.test.ts": 1415, - "src/main/ipc/worktrees-ssh-repo-owner-resolution.test.ts": 1447, "src/main/ipc/worktrees/listing/detected-provider-listing-catalog-version.test.ts": 650, "src/main/ipc/worktrees/listing/detected-worktree-scan-hygiene-gate.test.ts": 105, "src/main/ipc/worktrees/listing/detected-worktree-scan-listing-distro.test.ts": 880, diff --git a/config/scripts/orcad-distribution/server-bundle/session-scanner-service-smoke.test.mjs b/config/scripts/orcad-distribution/server-bundle/session-scanner-service-smoke.test.mjs deleted file mode 100644 index 60faa437f52..00000000000 --- a/config/scripts/orcad-distribution/server-bundle/session-scanner-service-smoke.test.mjs +++ /dev/null @@ -1,49 +0,0 @@ -import { build } from 'esbuild' -import { mkdtempSync, rmSync } from 'node:fs' -import { tmpdir } from 'node:os' -import { join, resolve } from 'node:path' -import { afterAll, beforeAll, describe, expect, it } from 'vitest' -import { smokeSessionScannerService } from './session-scanner-service-smoke.mjs' -import { - externalNativeAddons, - ORCAD_CHILD_ENTRY_POINTS, - ORCAD_EXTERNAL_MODULES, - orcadChildOutputFilename -} from './orcad-entry-build.mjs' - -const ENTRY = orcadChildOutputFilename(ORCAD_CHILD_ENTRY_POINTS.sessionScanner) -const directories = [] -let builtDirectory - -function fixtureDirectory() { - const directory = mkdtempSync(join(tmpdir(), 'orca-session-scanner-smoke-test-')) - directories.push(directory) - return directory -} - -beforeAll(async () => { - builtDirectory = fixtureDirectory() - await build({ - entryPoints: [resolve(ORCAD_CHILD_ENTRY_POINTS.sessionScanner)], - outfile: join(builtDirectory, ENTRY), - bundle: true, - platform: 'node', - target: 'node18', - format: 'cjs', - external: ORCAD_EXTERNAL_MODULES, - plugins: [externalNativeAddons], - logLevel: 'silent' - }) -}, 60_000) - -afterAll(() => { - for (const directory of directories) { - rmSync(directory, { recursive: true, force: true }) - } -}) - -describe('session scanner service build smoke', () => { - it('lists a seeded session through the built entry', () => { - expect(() => smokeSessionScannerService(builtDirectory)).not.toThrow() - }, 60_000) -}) diff --git a/src/main/ipc/worktrees-delete-pty-teardown.test.ts b/src/main/ipc/worktrees-delete-pty-teardown.test.ts deleted file mode 100644 index 0bf3532b539..00000000000 --- a/src/main/ipc/worktrees-delete-pty-teardown.test.ts +++ /dev/null @@ -1,192 +0,0 @@ -import type * as WorktreeScanCacheModule from '../git/worktrees/catalog/worktree-scan-cache' -import { beforeEach, describe, expect, it, vi } from 'vitest' -import { - listWorktreesMock, - removeWorktreeMock, - killAllProcessesForWorktreeMock -} from './worktrees-test-module-mocks' -import { handlers, setupWorktreeHandlers, store } from './worktrees-test-harness' -import { SshHostNotServedError } from '../providers/ssh-host-not-served-error' - -vi.mock('../git/worktrees/catalog/worktree-scan-cache', async (importOriginal) => { - return { - ...(await importOriginal()), - listWorktrees: (await import('./worktrees-test-module-mocks')).listWorktreesMock, - listWorktreesSharedStrictAllowingTrueEmpty: (await import('./worktrees-test-module-mocks')) - .listWorktreesMock, - listWorktreesSharedStrict: (await import('./worktrees-test-module-mocks')).listWorktreesMock - } -}) - -vi.mock('electron', async () => - (await import('./worktrees-test-module-mocks')).electronModuleMock() -) -vi.mock('../../shared/git/worktrees/git-worktree-porcelain-parser', async (importOriginal) => ({ - ...(await importOriginal>()), - parseWorktreeList: (await import('./worktrees-test-module-mocks')).parseWorktreeListMock -})) - -vi.mock('../git/worktrees/removal/worktree-removal-preflight', async (importOriginal) => ({ - ...(await importOriginal>()), - assertWorktreeCleanForRemoval: (await import('./worktrees-test-module-mocks')) - .assertWorktreeCleanForRemovalMock -})) - -vi.mock('../git/worktrees/creation/worktree-add', async (importOriginal) => ({ - ...(await importOriginal>()), - addWorktree: (await import('./worktrees-test-module-mocks')).addWorktreeMock -})) - -vi.mock('../git/worktrees/creation/worktree-sparse-add', async (importOriginal) => ({ - ...(await importOriginal>()), - addSparseWorktree: (await import('./worktrees-test-module-mocks')).addSparseWorktreeMock -})) - -vi.mock('../git/worktrees/removal/worktree-removal', async (importOriginal) => ({ - ...(await importOriginal>()), - removeWorktree: (await import('./worktrees-test-module-mocks')).removeWorktreeMock -})) - -vi.mock('../git/worktrees/removal/worktree-branch-removal', async (importOriginal) => ({ - ...(await importOriginal>()), - forceDeleteLocalBranch: (await import('./worktrees-test-module-mocks')).forceDeleteLocalBranchMock -})) -vi.mock('../git/worktrees/catalog/worktree-listing', async (importOriginal) => ({ - ...(await importOriginal>()), - listWorktreesStrict: (await import('./worktrees-test-module-mocks')).listWorktreesMock, - describeCreatedWorktree: (await import('./worktrees-test-module-mocks')) - .describeCreatedWorktreeMock -})) -const { getWorktreeRunnerMocks } = vi.hoisted(() => { - function createWorktreeRunnerMocks() { - return (async () => (await import('./worktrees-test-module-mocks')).gitRunnerModuleMock())() - } - let runner: ReturnType | undefined - return { getWorktreeRunnerMocks: () => (runner ??= createWorktreeRunnerMocks()) } -}) - -vi.mock('../git/command-runner/git-exec-file', async (importOriginal) => ({ - ...(await importOriginal>()), - gitExecFileAsync: (await getWorktreeRunnerMocks()).gitExecFileAsync -})) - -vi.mock('../git/command-runner/git-exec-file-sync', async (importOriginal) => ({ - ...(await importOriginal>()), - gitExecFileSync: (await getWorktreeRunnerMocks()).gitExecFileSync -})) -vi.mock('../git/remotes/repo-remote-queries', async () => ({ - getDefaultRemote: (await import('./worktrees-test-module-mocks')).getDefaultRemoteMock -})) -vi.mock('../git/refs/resolution/repo-branch-conflict', async () => - (await import('./worktrees-test-module-mocks')).gitBranchConflictModuleMock() -) - -vi.mock('../git/refs/resolution/repo-default-base-ref', async (importOriginal) => ({ - ...(await importOriginal>()), - getDefaultBaseRefAsync: (await import('./worktrees-test-module-mocks')).getBaseRefDefaultMock, - resolveDefaultBaseRefWithLocalGit: (await import('./worktrees-test-module-mocks')) - .resolveDefaultBaseRefWithLocalGitMock -})) -vi.mock('../git/repositories/git-username', async (importOriginal) => ({ - ...(await importOriginal>()), - resolveLocalGitUsername: (await import('./worktrees-test-module-mocks')) - .resolveLocalGitUsernameMock -})) -vi.mock('../github/client', async () => - (await import('./worktrees-test-module-mocks')).githubClientModuleMock() -) -vi.mock('../github/client/lookup/get-pr-for-branch', async () => - (await import('./worktrees-test-module-mocks')).githubClientModuleMock() -) -vi.mock('../source-control/hosted-review', async () => - (await import('./worktrees-test-module-mocks')).hostedReviewModuleMock() -) -vi.mock('./worktree-symlinks', async () => - (await import('./worktrees-test-module-mocks')).worktreeSymlinksModuleMock() -) - -vi.mock('../git/worktrees/content-paths/worktree-symlink-detection', async (importOriginal) => ({ - ...(await importOriginal>()), - findExistingWorktreeSymlinkPaths: (await import('./worktrees-test-module-mocks')) - .findExistingWorktreeSymlinkPathsMock -})) -vi.mock('../workspace-hooks/hooks', async () => - (await import('./worktrees-test-module-mocks')).hooksModuleMock() -) -vi.mock('../workspace-hooks/setup-runner-script-text', async (importOriginal) => - (await import('./worktrees-test-module-mocks')).setupRunnerScriptTextModuleMock( - (await importOriginal()) as Record - ) -) -vi.mock('../workspace-hooks/worktree-runner-script', async (importOriginal) => - (await import('./worktrees-test-module-mocks')).worktreeRunnerScriptModuleMock( - (await importOriginal()) as Record - ) -) -vi.mock('../workspace-hooks/effective-hook-config', async (importOriginal) => - (await import('./worktrees-test-module-mocks')).effectiveHookConfigModuleMock( - (await importOriginal()) as Record - ) -) -vi.mock('../workspace-hooks/setup-hook-env-vars', async (importOriginal) => - (await import('./worktrees-test-module-mocks')).setupHookEnvVarsModuleMock( - (await importOriginal()) as Record - ) -) -vi.mock('./worktree-logic', async (importOriginal) => - (await import('./worktrees-test-module-mocks')).worktreeLogicModuleMock( - (await importOriginal()) as Record - ) -) -vi.mock('../pty/history/terminal-history-deletion', async () => - (await import('./worktrees-test-module-mocks')).terminalHistoryDeletionModuleMock() -) -vi.mock('../ports/advertised-url-watcher', async () => - (await import('./worktrees-test-module-mocks')).advertisedUrlWatcherModuleMock() -) -vi.mock('../workspaces/cleanup/workspace-cleanup-scan-snapshot', async () => - (await import('./worktrees-test-module-mocks')).workspaceCleanupScanSnapshotModuleMock() -) -vi.mock('../workspaces/space-analysis/workspace-space-analysis-snapshot', async () => - (await import('./worktrees-test-module-mocks')).workspaceSpaceAnalysisSnapshotModuleMock() -) -vi.mock('../workspaces/cleanup/workspace-cleanup-removal-snapshot-prune', async () => - (await import('./worktrees-test-module-mocks')).workspaceCleanupRemovalSnapshotPruneModuleMock() -) -vi.mock('../runtime/workspaces/removal/worktree-teardown', async () => - (await import('./worktrees-test-module-mocks')).worktreeTeardownModuleMock() -) -vi.mock('./pty/provider/registry', async () => - (await import('./worktrees-test-module-mocks')).ptyRegistryModuleMock() -) -vi.mock('./pty/provider/state-cleanup', async () => - (await import('./worktrees-test-module-mocks')).ptyStateCleanupModuleMock() -) - -describe('registerWorktreeHandlers', () => { - beforeEach(() => { - setupWorktreeHandlers() - }) - - it('refuses an SSH worktree delete this process does not serve, never deleting here', async () => { - const repo = { - id: 'repo-ssh', - path: '/remote/repo', - displayName: 'ssh', - badgeColor: '#000', - addedAt: 0, - connectionId: 'conn-1', - worktreeBaseRef: null - } - store.getRepos.mockReturnValue([repo]) - store.getRepo.mockReturnValue(repo) - - await expect( - handlers['worktrees:remove'](null, { worktreeId: 'repo-ssh::/remote/feature-wt' }) - ).rejects.toThrow(SshHostNotServedError) - - expect(listWorktreesMock).not.toHaveBeenCalled() - expect(removeWorktreeMock).not.toHaveBeenCalled() - expect(killAllProcessesForWorktreeMock).not.toHaveBeenCalled() - }) -}) diff --git a/src/main/ipc/worktrees-issue-command-overrides.test.ts b/src/main/ipc/worktrees-issue-command-overrides.test.ts deleted file mode 100644 index 2526c4a954c..00000000000 --- a/src/main/ipc/worktrees-issue-command-overrides.test.ts +++ /dev/null @@ -1,217 +0,0 @@ -import type * as WorktreeScanCacheModule from '../git/worktrees/catalog/worktree-scan-cache' -import { beforeEach, describe, expect, it, vi } from 'vitest' -import { addWorktreeMock, getBranchConflictKindStrictMock } from './worktrees-test-module-mocks' -import { handlers, setupWorktreeHandlers, store } from './worktrees-test-harness' -import { SshHostNotServedError } from '../providers/ssh-host-not-served-error' - -vi.mock('../git/worktrees/catalog/worktree-scan-cache', async (importOriginal) => { - return { - ...(await importOriginal()), - listWorktrees: (await import('./worktrees-test-module-mocks')).listWorktreesMock, - listWorktreesSharedStrictAllowingTrueEmpty: (await import('./worktrees-test-module-mocks')) - .listWorktreesMock, - listWorktreesSharedStrict: (await import('./worktrees-test-module-mocks')).listWorktreesMock - } -}) - -vi.mock('electron', async () => - (await import('./worktrees-test-module-mocks')).electronModuleMock() -) -vi.mock('../../shared/git/worktrees/git-worktree-porcelain-parser', async (importOriginal) => ({ - ...(await importOriginal>()), - parseWorktreeList: (await import('./worktrees-test-module-mocks')).parseWorktreeListMock -})) - -vi.mock('../git/worktrees/removal/worktree-removal-preflight', async (importOriginal) => ({ - ...(await importOriginal>()), - assertWorktreeCleanForRemoval: (await import('./worktrees-test-module-mocks')) - .assertWorktreeCleanForRemovalMock -})) - -vi.mock('../git/worktrees/creation/worktree-add', async (importOriginal) => ({ - ...(await importOriginal>()), - addWorktree: (await import('./worktrees-test-module-mocks')).addWorktreeMock -})) - -vi.mock('../git/worktrees/creation/worktree-sparse-add', async (importOriginal) => ({ - ...(await importOriginal>()), - addSparseWorktree: (await import('./worktrees-test-module-mocks')).addSparseWorktreeMock -})) - -vi.mock('../git/worktrees/removal/worktree-removal', async (importOriginal) => ({ - ...(await importOriginal>()), - removeWorktree: (await import('./worktrees-test-module-mocks')).removeWorktreeMock -})) - -vi.mock('../git/worktrees/removal/worktree-branch-removal', async (importOriginal) => ({ - ...(await importOriginal>()), - forceDeleteLocalBranch: (await import('./worktrees-test-module-mocks')).forceDeleteLocalBranchMock -})) -vi.mock('../git/worktrees/catalog/worktree-listing', async (importOriginal) => ({ - ...(await importOriginal>()), - listWorktreesStrict: (await import('./worktrees-test-module-mocks')).listWorktreesMock, - describeCreatedWorktree: (await import('./worktrees-test-module-mocks')) - .describeCreatedWorktreeMock -})) -const { getWorktreeRunnerMocks } = vi.hoisted(() => { - function createWorktreeRunnerMocks() { - return (async () => (await import('./worktrees-test-module-mocks')).gitRunnerModuleMock())() - } - let runner: ReturnType | undefined - return { getWorktreeRunnerMocks: () => (runner ??= createWorktreeRunnerMocks()) } -}) - -vi.mock('../git/command-runner/git-exec-file', async (importOriginal) => ({ - ...(await importOriginal>()), - gitExecFileAsync: (await getWorktreeRunnerMocks()).gitExecFileAsync -})) - -vi.mock('../git/command-runner/git-exec-file-sync', async (importOriginal) => ({ - ...(await importOriginal>()), - gitExecFileSync: (await getWorktreeRunnerMocks()).gitExecFileSync -})) -vi.mock('../git/remotes/repo-remote-queries', async () => ({ - getDefaultRemote: (await import('./worktrees-test-module-mocks')).getDefaultRemoteMock -})) -vi.mock('../git/refs/resolution/repo-branch-conflict', async () => - (await import('./worktrees-test-module-mocks')).gitBranchConflictModuleMock() -) - -vi.mock('../git/refs/resolution/repo-default-base-ref', async (importOriginal) => ({ - ...(await importOriginal>()), - getDefaultBaseRefAsync: (await import('./worktrees-test-module-mocks')).getBaseRefDefaultMock, - resolveDefaultBaseRefWithLocalGit: (await import('./worktrees-test-module-mocks')) - .resolveDefaultBaseRefWithLocalGitMock -})) -vi.mock('../git/repositories/git-username', async (importOriginal) => ({ - ...(await importOriginal>()), - resolveLocalGitUsername: (await import('./worktrees-test-module-mocks')) - .resolveLocalGitUsernameMock -})) -vi.mock('../github/client', async () => - (await import('./worktrees-test-module-mocks')).githubClientModuleMock() -) -vi.mock('../github/client/lookup/get-pr-for-branch', async () => - (await import('./worktrees-test-module-mocks')).githubClientModuleMock() -) -vi.mock('../source-control/hosted-review', async () => - (await import('./worktrees-test-module-mocks')).hostedReviewModuleMock() -) -vi.mock('./worktree-symlinks', async () => - (await import('./worktrees-test-module-mocks')).worktreeSymlinksModuleMock() -) - -vi.mock('../git/worktrees/content-paths/worktree-symlink-detection', async (importOriginal) => ({ - ...(await importOriginal>()), - findExistingWorktreeSymlinkPaths: (await import('./worktrees-test-module-mocks')) - .findExistingWorktreeSymlinkPathsMock -})) -vi.mock('../workspace-hooks/hooks', async () => - (await import('./worktrees-test-module-mocks')).hooksModuleMock() -) -vi.mock('../workspace-hooks/setup-runner-script-text', async (importOriginal) => - (await import('./worktrees-test-module-mocks')).setupRunnerScriptTextModuleMock( - (await importOriginal()) as Record - ) -) -vi.mock('../workspace-hooks/worktree-runner-script', async (importOriginal) => - (await import('./worktrees-test-module-mocks')).worktreeRunnerScriptModuleMock( - (await importOriginal()) as Record - ) -) -vi.mock('../workspace-hooks/effective-hook-config', async (importOriginal) => - (await import('./worktrees-test-module-mocks')).effectiveHookConfigModuleMock( - (await importOriginal()) as Record - ) -) -vi.mock('../workspace-hooks/setup-hook-env-vars', async (importOriginal) => - (await import('./worktrees-test-module-mocks')).setupHookEnvVarsModuleMock( - (await importOriginal()) as Record - ) -) -vi.mock('./worktree-logic', async (importOriginal) => - (await import('./worktrees-test-module-mocks')).worktreeLogicModuleMock( - (await importOriginal()) as Record - ) -) -vi.mock('../pty/history/terminal-history-deletion', async () => - (await import('./worktrees-test-module-mocks')).terminalHistoryDeletionModuleMock() -) -vi.mock('../ports/advertised-url-watcher', async () => - (await import('./worktrees-test-module-mocks')).advertisedUrlWatcherModuleMock() -) -vi.mock('../workspaces/cleanup/workspace-cleanup-scan-snapshot', async () => - (await import('./worktrees-test-module-mocks')).workspaceCleanupScanSnapshotModuleMock() -) -vi.mock('../workspaces/space-analysis/workspace-space-analysis-snapshot', async () => - (await import('./worktrees-test-module-mocks')).workspaceSpaceAnalysisSnapshotModuleMock() -) -vi.mock('../workspaces/cleanup/workspace-cleanup-removal-snapshot-prune', async () => - (await import('./worktrees-test-module-mocks')).workspaceCleanupRemovalSnapshotPruneModuleMock() -) -vi.mock('../runtime/workspaces/removal/worktree-teardown', async () => - (await import('./worktrees-test-module-mocks')).worktreeTeardownModuleMock() -) -vi.mock('./pty/provider/registry', async () => - (await import('./worktrees-test-module-mocks')).ptyRegistryModuleMock() -) -vi.mock('./pty/provider/state-cleanup', async () => - (await import('./worktrees-test-module-mocks')).ptyStateCleanupModuleMock() -) - -const REMOTE_REPO_PATH = '/remote/repo' - -function makeRepo(fields: Record) { - return { - id: 'repo-1', - path: REMOTE_REPO_PATH, - displayName: 'repo', - badgeColor: '#000', - addedAt: 0, - worktreeBaseRef: 'origin/main', - ...fields - } -} - -function useRepo(repo: ReturnType): void { - store.getRepos.mockReturnValue([repo]) - store.getRepo.mockReturnValue(repo) - store.setWorktreeMeta.mockImplementation((_target: unknown, meta: unknown) => meta) -} - -describe('worktrees:create execution host routing', () => { - beforeEach(() => { - setupWorktreeHandlers() - }) - - it('refuses a row that names an SSH host only as executionHostId instead of creating locally', async () => { - useRepo(makeRepo({ executionHostId: 'ssh:target-a' })) - - await expect( - handlers['worktrees:create'](null, { repoId: 'repo-1', name: 'wt' }) - ).rejects.toThrow(SshHostNotServedError) - - expect(addWorktreeMock).not.toHaveBeenCalled() - }) -}) - -describe('worktrees:create exact name', () => { - beforeEach(() => { - setupWorktreeHandlers() - }) - - // Electron flattens the error to its message, so the renderer reads the code from the trailing token. - it('refuses a taken name with the code token instead of suffixing', async () => { - getBranchConflictKindStrictMock.mockResolvedValue('local') - - await expect( - handlers['worktrees:create'](null, { - repoId: 'repo-1', - name: 'Login fix', - branchIntent: 'create-exact' - }) - ).rejects.toThrow(/\nworktree_create_exact_name_taken$/) - - expect(addWorktreeMock).not.toHaveBeenCalled() - }) -}) diff --git a/src/main/ipc/worktrees-orphan-directory-cleanup.test.ts b/src/main/ipc/worktrees-orphan-directory-cleanup.test.ts index 47b3e4c734a..a587bb63165 100644 --- a/src/main/ipc/worktrees-orphan-directory-cleanup.test.ts +++ b/src/main/ipc/worktrees-orphan-directory-cleanup.test.ts @@ -228,45 +228,6 @@ describe('registerWorktreeHandlers', () => { } }) - it('prompts for force before removing an Orca-created orphaned worktree directory', async () => { - const parentDir = await mkdtemp(join(tmpdir(), 'orca-ipc-orphan-')) - const repoPath = join(parentDir, 'repo') - const orphanPath = join(parentDir, 'orphan') - const adminWorktreePath = join(repoPath, '.git', 'worktrees', 'orphan') - await mkdir(orphanPath, { recursive: true }) - await mkdir(adminWorktreePath, { recursive: true }) - await writeFile(join(orphanPath, '.git'), `gitdir: ${adminWorktreePath}\n`) - await writeFile(join(adminWorktreePath, 'gitdir'), `${join(orphanPath, '.git')}\n`) - const repo = { - id: 'repo-1', - path: repoPath, - displayName: 'repo', - badgeColor: '#000', - addedAt: 0, - worktreeBaseRef: null - } - store.getRepo.mockReturnValue(repo) - store.getRepos.mockReturnValue([repo]) - mockKnownFeatureWorktree(join(parentDir, 'real-feature'), repoPath) - store.getWorktreeMeta.mockReturnValue( - makeWorktreeMeta({ orcaCreatedAt: Date.now(), orcaCreationSource: 'runtime' }) - ) - - try { - await expect( - handlers['worktrees:remove'](null, { - worktreeId: `repo-1::${orphanPath}` - }) - ).rejects.toThrow('Worktree is no longer registered with Git but its directory remains.') - - await expect(lstat(orphanPath)).resolves.toBeTruthy() - expect(removeWorktreeMock).not.toHaveBeenCalled() - expect(store.removeWorktreeMeta).not.toHaveBeenCalled() - } finally { - await rm(parentDir, { recursive: true, force: true }) - } - }) - it('refuses an Orca-created unregistered directory with no git marker, even forced, until the user removes it', async () => { const parentDir = await mkdtemp(join(tmpdir(), 'orca-ipc-leftover-')) const repoPath = join(parentDir, 'repo') @@ -316,69 +277,4 @@ describe('registerWorktreeHandlers', () => { await rm(parentDir, { recursive: true, force: true }) } }) - - it('rejects an Orca-created unregistered local directory with a git directory', async () => { - const parentDir = await mkdtemp(join(tmpdir(), 'orca-ipc-standalone-')) - const repoPath = join(parentDir, 'repo') - const standalonePath = join(parentDir, 'standalone') - await mkdir(join(standalonePath, '.git'), { recursive: true }) - store.getRepo.mockReturnValue({ - id: 'repo-1', - path: repoPath, - displayName: 'repo', - badgeColor: '#000', - addedAt: 0, - worktreeBaseRef: null - }) - mockKnownFeatureWorktree(join(parentDir, 'real-feature'), repoPath) - store.getWorktreeMeta.mockReturnValue( - makeWorktreeMeta({ orcaCreatedAt: Date.now(), orcaCreationSource: 'runtime' }) - ) - - try { - await expect( - handlers['worktrees:remove'](null, { - worktreeId: `repo-1::${standalonePath}`, - force: true - }) - ).rejects.toThrow(`Refusing to delete unregistered worktree path: ${standalonePath}`) - - await expect(lstat(standalonePath)).resolves.toBeTruthy() - expect(removeWorktreeMock).not.toHaveBeenCalled() - expect(store.removeWorktreeMeta).not.toHaveBeenCalled() - } finally { - await rm(parentDir, { recursive: true, force: true }) - } - }) - - it('does not inspect or delete a local path for an SSH orphan this process does not serve', async () => { - const localPath = await mkdtemp(join(tmpdir(), 'orca-ipc-ssh-missing-fs-')) - const repo = { - id: 'repo-ssh-missing-fs', - path: '/remote/repo', - displayName: 'ssh', - badgeColor: '#000', - addedAt: 0, - connectionId: 'conn-missing-fs' - } - store.getRepo.mockReturnValue(repo) - store.getWorktreeMeta.mockReturnValue( - makeWorktreeMeta({ orcaCreatedAt: Date.now(), orcaCreationSource: 'ssh' }) - ) - - try { - await expect( - handlers['worktrees:remove'](null, { - worktreeId: `${repo.id}::${localPath}`, - force: true - }) - ).rejects.toThrow('isn’t served by Orca’s server') - - await expect(lstat(localPath)).resolves.toBeTruthy() - expect(removeWorktreeMock).not.toHaveBeenCalled() - expect(store.removeWorktreeMeta).not.toHaveBeenCalled() - } finally { - await rm(localPath, { recursive: true, force: true }) - } - }) }) diff --git a/src/main/ipc/worktrees-ssh-repo-owner-resolution.test.ts b/src/main/ipc/worktrees-ssh-repo-owner-resolution.test.ts deleted file mode 100644 index 79baa10474e..00000000000 --- a/src/main/ipc/worktrees-ssh-repo-owner-resolution.test.ts +++ /dev/null @@ -1,260 +0,0 @@ -import type * as WorktreeScanCacheModule from '../git/worktrees/catalog/worktree-scan-cache' -import { beforeEach, describe, expect, it, vi } from 'vitest' -import type { ProviderRequestId } from '../../shared/detected-worktree-provider-contract' -import { toSshExecutionHostId } from '../../shared/execution-host' -import { getSshProviderAuthority } from '../ssh/connection/authority/ssh-provider-authority' -import { listWorktreesMock } from './worktrees-test-module-mocks' -import { handlers, ipcEvent, setupWorktreeHandlers, store } from './worktrees-test-harness' - -vi.mock('../git/worktrees/catalog/worktree-scan-cache', async (importOriginal) => { - return { - ...(await importOriginal()), - listWorktrees: (await import('./worktrees-test-module-mocks')).listWorktreesMock, - listWorktreesSharedStrictAllowingTrueEmpty: (await import('./worktrees-test-module-mocks')) - .listWorktreesMock, - listWorktreesSharedStrict: (await import('./worktrees-test-module-mocks')).listWorktreesMock - } -}) - -vi.mock('electron', async () => - (await import('./worktrees-test-module-mocks')).electronModuleMock() -) -vi.mock('../../shared/git/worktrees/git-worktree-porcelain-parser', async (importOriginal) => ({ - ...(await importOriginal>()), - parseWorktreeList: (await import('./worktrees-test-module-mocks')).parseWorktreeListMock -})) - -vi.mock('../git/worktrees/removal/worktree-removal-preflight', async (importOriginal) => ({ - ...(await importOriginal>()), - assertWorktreeCleanForRemoval: (await import('./worktrees-test-module-mocks')) - .assertWorktreeCleanForRemovalMock -})) - -vi.mock('../git/worktrees/creation/worktree-add', async (importOriginal) => ({ - ...(await importOriginal>()), - addWorktree: (await import('./worktrees-test-module-mocks')).addWorktreeMock -})) - -vi.mock('../git/worktrees/creation/worktree-sparse-add', async (importOriginal) => ({ - ...(await importOriginal>()), - addSparseWorktree: (await import('./worktrees-test-module-mocks')).addSparseWorktreeMock -})) - -vi.mock('../git/worktrees/removal/worktree-removal', async (importOriginal) => ({ - ...(await importOriginal>()), - removeWorktree: (await import('./worktrees-test-module-mocks')).removeWorktreeMock -})) - -vi.mock('../git/worktrees/removal/worktree-branch-removal', async (importOriginal) => ({ - ...(await importOriginal>()), - forceDeleteLocalBranch: (await import('./worktrees-test-module-mocks')).forceDeleteLocalBranchMock -})) -vi.mock('../git/worktrees/catalog/worktree-listing', async (importOriginal) => ({ - ...(await importOriginal>()), - listWorktreesStrict: (await import('./worktrees-test-module-mocks')).listWorktreesMock, - describeCreatedWorktree: (await import('./worktrees-test-module-mocks')) - .describeCreatedWorktreeMock -})) -const { getWorktreeRunnerMocks } = vi.hoisted(() => { - function createWorktreeRunnerMocks() { - return (async () => (await import('./worktrees-test-module-mocks')).gitRunnerModuleMock())() - } - let runner: ReturnType | undefined - return { getWorktreeRunnerMocks: () => (runner ??= createWorktreeRunnerMocks()) } -}) - -vi.mock('../git/command-runner/git-exec-file', async (importOriginal) => ({ - ...(await importOriginal>()), - gitExecFileAsync: (await getWorktreeRunnerMocks()).gitExecFileAsync -})) - -vi.mock('../git/command-runner/git-exec-file-sync', async (importOriginal) => ({ - ...(await importOriginal>()), - gitExecFileSync: (await getWorktreeRunnerMocks()).gitExecFileSync -})) -vi.mock('../git/remotes/repo-remote-queries', async () => ({ - getDefaultRemote: (await import('./worktrees-test-module-mocks')).getDefaultRemoteMock -})) -vi.mock('../git/refs/resolution/repo-branch-conflict', async () => - (await import('./worktrees-test-module-mocks')).gitBranchConflictModuleMock() -) - -vi.mock('../git/refs/resolution/repo-default-base-ref', async (importOriginal) => ({ - ...(await importOriginal>()), - getDefaultBaseRefAsync: (await import('./worktrees-test-module-mocks')).getBaseRefDefaultMock, - resolveDefaultBaseRefWithLocalGit: (await import('./worktrees-test-module-mocks')) - .resolveDefaultBaseRefWithLocalGitMock -})) -vi.mock('../git/repositories/git-username', async (importOriginal) => ({ - ...(await importOriginal>()), - resolveLocalGitUsername: (await import('./worktrees-test-module-mocks')) - .resolveLocalGitUsernameMock -})) -vi.mock('../github/client', async () => - (await import('./worktrees-test-module-mocks')).githubClientModuleMock() -) -vi.mock('../github/client/lookup/get-pr-for-branch', async () => - (await import('./worktrees-test-module-mocks')).githubClientModuleMock() -) -vi.mock('../source-control/hosted-review', async () => - (await import('./worktrees-test-module-mocks')).hostedReviewModuleMock() -) -vi.mock('./worktree-symlinks', async () => - (await import('./worktrees-test-module-mocks')).worktreeSymlinksModuleMock() -) - -vi.mock('../git/worktrees/content-paths/worktree-symlink-detection', async (importOriginal) => ({ - ...(await importOriginal>()), - findExistingWorktreeSymlinkPaths: (await import('./worktrees-test-module-mocks')) - .findExistingWorktreeSymlinkPathsMock -})) -vi.mock('../workspace-hooks/hooks', async () => - (await import('./worktrees-test-module-mocks')).hooksModuleMock() -) -vi.mock('../workspace-hooks/setup-runner-script-text', async (importOriginal) => - (await import('./worktrees-test-module-mocks')).setupRunnerScriptTextModuleMock( - (await importOriginal()) as Record - ) -) -vi.mock('../workspace-hooks/worktree-runner-script', async (importOriginal) => - (await import('./worktrees-test-module-mocks')).worktreeRunnerScriptModuleMock( - (await importOriginal()) as Record - ) -) -vi.mock('../workspace-hooks/effective-hook-config', async (importOriginal) => - (await import('./worktrees-test-module-mocks')).effectiveHookConfigModuleMock( - (await importOriginal()) as Record - ) -) -vi.mock('../workspace-hooks/setup-hook-env-vars', async (importOriginal) => - (await import('./worktrees-test-module-mocks')).setupHookEnvVarsModuleMock( - (await importOriginal()) as Record - ) -) -vi.mock('./worktree-logic', async (importOriginal) => - (await import('./worktrees-test-module-mocks')).worktreeLogicModuleMock( - (await importOriginal()) as Record - ) -) -vi.mock('../pty/history/terminal-history-deletion', async () => - (await import('./worktrees-test-module-mocks')).terminalHistoryDeletionModuleMock() -) -vi.mock('../ports/advertised-url-watcher', async () => - (await import('./worktrees-test-module-mocks')).advertisedUrlWatcherModuleMock() -) -vi.mock('../workspaces/cleanup/workspace-cleanup-scan-snapshot', async () => - (await import('./worktrees-test-module-mocks')).workspaceCleanupScanSnapshotModuleMock() -) -vi.mock('../workspaces/space-analysis/workspace-space-analysis-snapshot', async () => - (await import('./worktrees-test-module-mocks')).workspaceSpaceAnalysisSnapshotModuleMock() -) -vi.mock('../workspaces/cleanup/workspace-cleanup-removal-snapshot-prune', async () => - (await import('./worktrees-test-module-mocks')).workspaceCleanupRemovalSnapshotPruneModuleMock() -) -vi.mock('../runtime/workspaces/removal/worktree-teardown', async () => - (await import('./worktrees-test-module-mocks')).worktreeTeardownModuleMock() -) -vi.mock('./pty/provider/registry', async () => - (await import('./worktrees-test-module-mocks')).ptyRegistryModuleMock() -) -vi.mock('./pty/provider/state-cleanup', async () => - (await import('./worktrees-test-module-mocks')).ptyStateCleanupModuleMock() -) - -describe('registerWorktreeHandlers', () => { - beforeEach(() => { - setupWorktreeHandlers() - }) - - // Runtime rows bypass local stale GC too, so only authoritative host scans retire them (#17776). - - // Runtime ownership plus a connectionId is contradictory; findExactRepoOwner must refuse it. - - it('fails closed for duplicate exact owners and ambiguous legacy repo IDs', async () => { - const sshRepo = { - id: 'shared-repo', - path: '/remote/repo-a', - displayName: 'remote repo', - badgeColor: '#000', - addedAt: 0, - connectionId: 'target-a' - } - const duplicateSshRepo = { ...sshRepo, path: '/remote/repo-b' } - const localRepo = { ...sshRepo, path: '/local/repo', connectionId: undefined } - store.getRepos.mockReturnValue([sshRepo, duplicateSshRepo, localRepo]) - const expectedAuthority = getSshProviderAuthority('target-a') - - const qualified = await handlers['worktrees:listDetected'](ipcEvent, { - providerRequestId: 'request-1' as ProviderRequestId, - repoId: 'shared-repo', - executionHostId: toSshExecutionHostId('target-a'), - expectedAuthority - }) - const legacy = await handlers['worktrees:listDetected'](null, { repoId: 'shared-repo' }) - - expect(qualified).toMatchObject({ - status: 'ambiguous-owner', - providerRequestId: 'request-1', - executionHostId: 'ssh:target-a' - }) - expect(legacy).toEqual({ - repoId: 'shared-repo', - authoritative: false, - source: 'metadata-fallback', - worktrees: [] - }) - expect(listWorktreesMock).not.toHaveBeenCalled() - }) - - it('rejects runtime hosts, wrong SSH targets, and missing authority', async () => { - const requestId = 'request-1' as ProviderRequestId - const expectedAuthority = getSshProviderAuthority('target-a') - - const runtimeResult = await handlers['worktrees:listDetected'](ipcEvent, { - providerRequestId: requestId, - repoId: 'repo-1', - executionHostId: 'runtime:runtime-a' - }) - const wrongTargetResult = await handlers['worktrees:listDetected'](ipcEvent, { - providerRequestId: requestId, - repoId: 'repo-1', - executionHostId: toSshExecutionHostId('target-b'), - expectedAuthority - }) - const missingAuthorityResult = await handlers['worktrees:listDetected'](ipcEvent, { - providerRequestId: requestId, - repoId: 'repo-1', - executionHostId: toSshExecutionHostId('target-a') - }) - const zeroOwnerResult = await handlers['worktrees:listDetected'](ipcEvent, { - providerRequestId: requestId, - repoId: 'repo-1', - executionHostId: toSshExecutionHostId('target-a'), - expectedAuthority - }) - store.getRepos.mockReturnValue([ - { - id: 'repo-1', - path: '/remote/repo', - displayName: 'repo', - badgeColor: '#000', - addedAt: 0, - connectionId: 'target-b', - executionHostId: toSshExecutionHostId('target-a') - } - ]) - const wrongProviderOwnerResult = await handlers['worktrees:listDetected'](ipcEvent, { - providerRequestId: requestId, - repoId: 'repo-1', - executionHostId: toSshExecutionHostId('target-a'), - expectedAuthority - }) - - expect(runtimeResult).toMatchObject({ status: 'rejected' }) - expect(wrongTargetResult).toMatchObject({ status: 'rejected' }) - expect(missingAuthorityResult).toMatchObject({ status: 'rejected' }) - expect(zeroOwnerResult).toMatchObject({ status: 'ambiguous-owner' }) - expect(wrongProviderOwnerResult).toMatchObject({ status: 'rejected' }) - expect(listWorktreesMock).not.toHaveBeenCalled() - }) -}) diff --git a/src/main/native-chat/agent-session-wire/options/testing/at-rest-command-catalog-stamp.test.ts b/src/main/native-chat/agent-session-wire/options/testing/at-rest-command-catalog-stamp.test.ts index dad95873f13..5f5cf711ccc 100644 --- a/src/main/native-chat/agent-session-wire/options/testing/at-rest-command-catalog-stamp.test.ts +++ b/src/main/native-chat/agent-session-wire/options/testing/at-rest-command-catalog-stamp.test.ts @@ -6,14 +6,12 @@ import { AtRestCommandCatalog } from '../at-rest-command-catalog' function catalogWithSkills() { let stamp = 'a' let skills = ['old-skill'] - let scans = 0 const catalog = new AtRestCommandCatalog({ label: 'test', accepts: () => true, resolveWorkspacePath: async () => tmpdir(), stamp: async () => stamp, scan: async () => { - scans += 1 return { commands: skills.map((name) => ({ name, kind: 'skill' as const, path: `/s/${name}` })), fromAgent: true @@ -22,7 +20,6 @@ function catalogWithSkills() { }) return { catalog, - scans: () => scans, install: (name: string, nextStamp: string) => { skills = [...skills, name] stamp = nextStamp @@ -34,13 +31,12 @@ const names = (commands: { name: string }[] | undefined) => commands?.map(({ nam describe('at-rest commands before the TTL', () => { it('rescans on a menu open once the skill sources changed', async () => { - const { catalog, scans, install } = catalogWithSkills() + const { catalog, install } = catalogWithSkills() const record = hostTestRestingOptionsRecord() expect(names(await catalog.settled(record, 1_000, true))).toEqual(['old-skill']) install('perf-scan', 'b') expect(names(await catalog.settled(record, 1_000, true))).toEqual(['old-skill', 'perf-scan']) - expect(scans()).toBe(2) catalog.dispose() }) }) diff --git a/src/main/native-chat/structured-chat-naming-title.test.ts b/src/main/native-chat/structured-chat-naming-title.test.ts index fbc58ba7566..48d8da27fc3 100644 --- a/src/main/native-chat/structured-chat-naming-title.test.ts +++ b/src/main/native-chat/structured-chat-naming-title.test.ts @@ -70,28 +70,6 @@ function rig( } } -it('tries CLI reuse once while independent generation retries, and skips it on manual regeneration', async () => { - const reuse = vi.fn() - const generate = vi.fn( - async (_record, _prompt, _current, automatic) => { - if (automatic) { - reuse() - } - return generate.mock.calls.length === 1 - ? { error: 'Retry independent generation' } - : { name: 'Login repair' } - } - ) - const state = rig(generate) - state.start() - await state.done() - expect(state.record().conversationName).toBe('Login repair') - expect(reuse).toHaveBeenCalledOnce() - state.handler.regenerate('chat-1') - await state.done() - expect(reuse).toHaveBeenCalledOnce() -}) - it.each(['owner', 'same-text rename'] as const)( 'rejects a queued title write after %s changes', async (change) => { diff --git a/src/main/opencode/model-launch/opencode-root-model-overlay.test.ts b/src/main/opencode/model-launch/opencode-root-model-overlay.test.ts index abe8099fd87..c010d63ab9c 100644 --- a/src/main/opencode/model-launch/opencode-root-model-overlay.test.ts +++ b/src/main/opencode/model-launch/opencode-root-model-overlay.test.ts @@ -21,15 +21,6 @@ describe('planOpenCodeRootModelRewrite', () => { ).toBe(`& opencode --standalone${tail}`) }) - it('uses the last of repeated models and an existing --standalone', () => { - expect( - planOpenCodeRootModelRewrite( - `opencode2 --standalone ${q(['--model', 'a/b', '--model=c/d'], 'posix')}`, - 'posix' - ) - ).toMatchObject({ command: 'opencode2 --standalone', model: 'c/d' }) - }) - it.each([ ['posix', `opencode ${q(['run', '--model', 'a/b'], 'posix')}`], ['posix', `opencode ${q(['./pkg', '--model', 'a/b'], 'posix')}`], @@ -51,13 +42,4 @@ describe('planOpenCodeRootModelRewrite', () => { ] as const)('leaves %s `%s` as typed', (shell, command) => { expect(planOpenCodeRootModelRewrite(command, shell, 'C:\\')).toBeNull() }) - - it('probes the primary agent only for a new session on it', () => { - expect( - planOpenCodeRootModelRewrite( - `opencode ${q(['-s', 'ses_1', '--model', 'a/b'], 'posix')}`, - 'posix' - )?.probeExecutable - ).toBeUndefined() - }) }) From 7c794af26dd5e4cbaa2bdf99372809bcbb188834 Mon Sep 17 00:00:00 2001 From: Kelvin Amoaba <97001695+AmoabaKelvin@users.noreply.github.com> Date: Thu, 12 Nov 2026 03:04:09 +0000 Subject: [PATCH 06/13] fix(speech): commit dictation once per span, not at every pause (#41297) Each pause cut was decoded and committed on its own, so Parakeet capitalized the word after a mid-sentence pause. Pause cuts now only update the live text. --- src/main/speech/stt-recognizer.ts | 8 +- src/main/speech/stt-service.ts | 2 +- .../speech/stt-transcribe-cpp-recognizer.ts | 99 +++++++++++++++---- src/main/speech/stt-vad-pause-segmenter.ts | 2 +- 4 files changed, 86 insertions(+), 25 deletions(-) diff --git a/src/main/speech/stt-recognizer.ts b/src/main/speech/stt-recognizer.ts index fb440a87295..667e3d6d719 100644 --- a/src/main/speech/stt-recognizer.ts +++ b/src/main/speech/stt-recognizer.ts @@ -6,14 +6,16 @@ export type SttRecognizerInit = { sampleRate: number /** Absent = auto-detect. */ language?: string - /** Silero VAD GGUF; when present, offline models commit text at each pause. */ + /** Silero VAD GGUF; when present, offline models show live text at each pause. */ vadModelPath?: string } export type SttRecognizerEvent = - /** One decoded offline chunk; `samples` counts the stream's input-rate samples decoded so far. */ + /** One committed offline chunk; `samples` counts the stream's input-rate samples committed so far. */ | { type: 'chunk'; streamId: string; text: string; samples: number } - /** Pause-cut streams only: speech was heard that no chunk has committed yet. */ + /** The uncommitted phrases' live text so far; the next chunk replaces it. */ + | { type: 'partial'; streamId: string; text: string } + /** Pause-cut streams only: speech was heard whose text hasn't arrived yet. */ | { type: 'speech-pending'; streamId: string; pending: boolean } /** A load step finished, so a slow load is told apart from a hung GPU. */ | { type: 'loading' } diff --git a/src/main/speech/stt-service.ts b/src/main/speech/stt-service.ts index bacc4b753ed..97231babfbc 100644 --- a/src/main/speech/stt-service.ts +++ b/src/main/speech/stt-service.ts @@ -24,7 +24,7 @@ export type SttEvent = | { type: 'ready' } | { type: 'partial'; text?: string } | { type: 'final'; text?: string } - /** Speech was heard that no final has committed yet; only from pause-cut on-device models. */ + /** Speech was heard whose text hasn't arrived yet; only from pause-cut on-device models. */ | { type: 'speech-pending'; pending: boolean } | { type: 'stopped' } | { type: 'error'; error?: string } diff --git a/src/main/speech/stt-transcribe-cpp-recognizer.ts b/src/main/speech/stt-transcribe-cpp-recognizer.ts index f292db59403..26b3a793039 100644 --- a/src/main/speech/stt-transcribe-cpp-recognizer.ts +++ b/src/main/speech/stt-transcribe-cpp-recognizer.ts @@ -2,8 +2,9 @@ import { availableParallelism } from 'node:os' import { join } from 'node:path' import { pathToFileURL } from 'node:url' import type * as TranscribeCpp from 'transcribe-cpp' +import { appendDictationText } from '../../shared/dictation-transcript-spacing' import { resampleToRate } from './stt-audio-resample' -import { OfflineAudioChunker } from './stt-offline-audio-chunker' +import { OFFLINE_DECODE_CHUNK_SECONDS, OfflineAudioChunker } from './stt-offline-audio-chunker' import { VadPauseSegmenter } from './stt-vad-pause-segmenter' import type { SttRecognizer, SttRecognizerInit, SttRecognizerPost } from './stt-recognizer' @@ -44,8 +45,12 @@ type StreamCoverage = { inputSamples: number resampledSamples: number reportedSamples: number + /** Stream position the last phrase decode reached, for its speed check. */ + decodedSamples: number slowPhraseDecodes: number speechPending: boolean + /** Phrases since the last commit and their live text. */ + span: { audio: Float32Array[]; samples: number; texts: string[] } } // Whisper pads every decode to a 30 s window, so each phrase costs a full decode. Where that takes longer than the @@ -56,6 +61,9 @@ const SLOW_PHRASE_DECODES_BEFORE_NO_CUTS = 2 // Dropped silence still counts as covered; report it so the host's replay buffer and stall watch don't grow through long pauses. const SILENCE_COVERAGE_REPORT_SECONDS = 5 +// The CPU fallback keeps uncommitted audio for a replay and takes over 30 s of it for a stalled GPU. +const COMMIT_AFTER_SECONDS = 20 + // Optional: without it dictation still works, with text only at the decode bound and at stop. async function loadSileroVad( transcribe: typeof TranscribeCpp, @@ -128,8 +136,10 @@ export async function createTranscribeCppRecognizer( inputSamples: 0, resampledSamples: 0, reportedSamples: 0, + decodedSamples: 0, slowPhraseDecodes: 0, - speechPending: false + speechPending: false, + span: { audio: [], samples: 0, texts: [] } } if (vad && vadModel) { stream.segmenter = new VadPauseSegmenter( @@ -162,9 +172,13 @@ export async function createTranscribeCppRecognizer( // The first run compiles GPU pipelines (seconds on Metal); pay it while loading, not on the first stop. await transcribeChunk(new Float32Array(Math.round(modelRate / 2)), baseOptions) + // Stream position (model-rate samples) the segmenter has handed on, less chunks still waiting to be decoded. + const position = (stream: StreamCoverage, undecoded: number) => + stream.resampledSamples - stream.segmenter.heldSamples - undecoded + // Coverage is in input-rate samples so the host can trim the audio it keeps for a CPU replay. const report = (streamId: string, stream: StreamCoverage, text: string, undecoded: number) => { - const covered = stream.resampledSamples - stream.segmenter.heldSamples - undecoded + const covered = position(stream, undecoded) stream.reportedSamples = covered post({ type: 'chunk', @@ -177,28 +191,59 @@ export async function createTranscribeCppRecognizer( // Canceled streams, until their queued discard runs; also covers a finish that already closed its stream. const canceled = new Set() - const decode = async (streamId: string, stream: StreamCoverage, chunks: Float32Array[]) => { + const spanLimit = OFFLINE_DECODE_CHUNK_SECONDS * modelRate + + // The span's phrases decode again together, so a pause mid-sentence doesn't capitalize the words after it. + const commit = async (streamId: string, stream: StreamCoverage, undecoded: number) => { + const { audio, texts } = stream.span + if (audio.length === 0 || canceled.has(streamId)) { + return + } + stream.span = { audio: [], samples: 0, texts: [] } + const text = + audio.length === 1 && texts.length === 1 + ? texts[0] + : (await transcribeChunk(joinAudio(audio), stream.options)).trim() + report(streamId, stream, text, undecoded) + } + + /** `live`: decode each phrase as it ends for live text; at stop only the commit decodes. */ + const decode = async ( + streamId: string, + stream: StreamCoverage, + chunks: Float32Array[], + live: boolean + ) => { let undecoded = chunks.reduce((total, chunk) => total + chunk.length, 0) for (const chunk of chunks) { if (canceled.has(streamId)) { return } - const startedAt = performance.now() - const text = await transcribeChunk(chunk, stream.options) - const decodeMs = performance.now() - startedAt - const spokenSince = stream.reportedSamples - undecoded -= chunk.length - report(streamId, stream, text.trim(), undecoded) - if (stream.vad) { - noteDecodeSpeed( - stream, - decodeMs, - ((stream.reportedSamples - spokenSince) / modelRate) * 1000 - ) + if (stream.span.samples + chunk.length > spanLimit) { + await commit(streamId, stream, undecoded) } + undecoded -= chunk.length + stream.span.audio.push(chunk) + stream.span.samples += chunk.length + if (!live) { + continue + } + const startedAt = performance.now() + const text = (await transcribeChunk(chunk, stream.options)).trim() + const decodeMs = performance.now() - startedAt + stream.span.texts.push(text) + post({ type: 'partial', streamId, text: stream.span.texts.reduce(appendDictationText, '') }) + const reached = position(stream, undecoded) + if (stream.vad) { + noteDecodeSpeed(stream, decodeMs, ((reached - stream.decodedSamples) / modelRate) * 1000) + } + stream.decodedSamples = reached } - const dropped = stream.resampledSamples - stream.segmenter.heldSamples - stream.reportedSamples - if (dropped >= SILENCE_COVERAGE_REPORT_SECONDS * modelRate) { + if (stream.resampledSamples - stream.reportedSamples >= COMMIT_AFTER_SECONDS * modelRate) { + await commit(streamId, stream, 0) + } + const dropped = position(stream, 0) - stream.reportedSamples + if (stream.span.samples === 0 && dropped >= SILENCE_COVERAGE_REPORT_SECONDS * modelRate) { report(streamId, stream, '', 0) } } @@ -244,7 +289,7 @@ export async function createTranscribeCppRecognizer( const phrases = await stream.segmenter.push(resampled) // A cut phrase stays pending until its text is posted. notePendingSpeech(streamId, stream, phrases.length > 0 || !!stream.segmenter.phraseOpen) - await decode(streamId, stream, phrases) + await decode(streamId, stream, phrases, true) notePendingSpeech(streamId, stream, !!stream.segmenter.phraseOpen) }, async finish(streamId) { @@ -253,7 +298,8 @@ export async function createTranscribeCppRecognizer( return } try { - await decode(streamId, stream, await stream.segmenter.flush()) + await decode(streamId, stream, await stream.segmenter.flush(), false) + await commit(streamId, stream, 0) notePendingSpeech(streamId, stream, false) } finally { stream.vad?.dispose() @@ -268,3 +314,16 @@ export async function createTranscribeCppRecognizer( } } } + +function joinAudio(parts: Float32Array[]): Float32Array { + if (parts.length === 1) { + return parts[0] + } + const joined = new Float32Array(parts.reduce((total, part) => total + part.length, 0)) + let offset = 0 + for (const part of parts) { + joined.set(part, offset) + offset += part.length + } + return joined +} diff --git a/src/main/speech/stt-vad-pause-segmenter.ts b/src/main/speech/stt-vad-pause-segmenter.ts index 8e7fc3b3d33..04af6067957 100644 --- a/src/main/speech/stt-vad-pause-segmenter.ts +++ b/src/main/speech/stt-vad-pause-segmenter.ts @@ -4,7 +4,7 @@ import { OfflineAudioChunker } from './stt-offline-audio-chunker' // Silero's own defaults: a frame at or above this is speech; inside speech, one below the second is silence. const SPEECH_PROBABILITY = 0.5 const SILENCE_PROBABILITY = 0.35 -// A pause this long commits the phrase before it. Shorter cuts split phrases, and each piece gets its own capital and full stop. +// A pause this long ends the phrase before it, which then decodes for live text; every cut costs a decode. export const VAD_PAUSE_SECONDS = 0.6 // Silence kept ahead of detected speech so a soft first syllable isn't clipped. const SPEECH_PREROLL_SECONDS = 0.4 From 129227686b9edce81089b3c51d33e8c2a14efd5c Mon Sep 17 00:00:00 2001 From: Kelvin Amoaba <97001695+AmoabaKelvin@users.noreply.github.com> Date: Thu, 12 Nov 2026 03:06:04 +0000 Subject: [PATCH 07/13] fix(editor): open a wikilink from the right half of its label (#41337) ProseMirror resolves a click on the right half of an inline atom to the position after it, so the rich editor's modifier-click router read the following node. Wikilinks and superscript citation links did nothing when clicked there. Resolve the atom the click landed on instead. Co-authored-by: Wooseong Kim --- ...rich-markdown-editor-click-routing.test.ts | 79 +++++++++++++++++++ .../rich-markdown-editor-click-routing.ts | 16 +++- 2 files changed, 94 insertions(+), 1 deletion(-) create mode 100644 src/renderer/src/components/editor/rich-markdown/links/rich-markdown-editor-click-routing.test.ts diff --git a/src/renderer/src/components/editor/rich-markdown/links/rich-markdown-editor-click-routing.test.ts b/src/renderer/src/components/editor/rich-markdown/links/rich-markdown-editor-click-routing.test.ts new file mode 100644 index 00000000000..b9155234366 --- /dev/null +++ b/src/renderer/src/components/editor/rich-markdown/links/rich-markdown-editor-click-routing.test.ts @@ -0,0 +1,79 @@ +// @vitest-environment happy-dom + +import { Editor } from '@tiptap/core' +import { afterEach, describe, expect, it } from 'vitest' +import { createRichMarkdownHtmlSuperscriptLinkContext } from '../citations/rich-markdown-html-superscript-link-context' +import { createRichMarkdownEditorCodec } from '../document/rich-markdown-source-transport' +import { encodeRawMarkdownHtmlForRichEditor } from '../raw-source/raw-markdown-html' +import { createRichMarkdownExtensions } from '../schema/rich-markdown-extensions' +import { handleRichMarkdownEditorClick } from './rich-markdown-editor-click-routing' + +const editors: Editor[] = [] + +afterEach(() => { + for (const editor of editors.splice(0)) { + editor.destroy() + } +}) + +function createDocLinkEditor(): Editor { + const codec = createRichMarkdownEditorCodec() + const editor = new Editor({ + element: document.body.appendChild(document.createElement('div')), + extensions: createRichMarkdownExtensions({ codec }), + content: encodeRawMarkdownHtmlForRichEditor('See [[docs/guide|some label]] next', codec), + contentType: 'markdown' + }) + editors.push(editor) + return editor +} + +function docLinkEnd(editor: Editor): number { + let end = -1 + editor.state.doc.descendants((node, pos) => { + if (node.type.name === 'markdownDocLink') { + end = pos + node.nodeSize + } + }) + return end +} + +// posAtCoords puts a click on the right half of an inline atom at the position after it. +function modClickAt(editor: Editor, pos: number, target: Element): string[] { + const opened: string[] = [] + const event = new MouseEvent('click', { metaKey: true }) + Object.defineProperty(event, 'target', { value: target }) + handleRichMarkdownEditorClick({ + activateMarkdownLink: () => undefined, + editorRef: { current: editor }, + event, + filePath: '/repo/docs/index.md', + isMac: true, + htmlSuperscriptLinkContext: createRichMarkdownHtmlSuperscriptLinkContext({ + sourceFilePath: '/repo/docs/index.md', + worktreeId: 'worktree-1', + worktreeRoot: '/repo', + sourceOwner: { kind: 'local' } + }), + markdownCommentsRef: { current: [] }, + markdownSourceLineOffsetRef: { current: 0 }, + onOpenDocLinkRef: { current: (docTarget) => opened.push(docTarget) }, + pos, + rootRef: { current: null }, + scrollRichMarkdownReviewNoteCardIntoView: () => undefined, + view: editor.view, + worktreeId: 'worktree-1', + worktreeRoot: '/repo' + }) + return opened +} + +describe('rich markdown modifier-click on a wikilink', () => { + it('opens the wikilink from the right half of its label', () => { + const editor = createDocLinkEditor() + const label = editor.view.dom.querySelector('[data-doc-link-target]') + expect(label).not.toBeNull() + + expect(modClickAt(editor, docLinkEnd(editor), label!)).toEqual(['docs/guide']) + }) +}) diff --git a/src/renderer/src/components/editor/rich-markdown/links/rich-markdown-editor-click-routing.ts b/src/renderer/src/components/editor/rich-markdown/links/rich-markdown-editor-click-routing.ts index bf4e6a7d2bc..41a5000de79 100644 --- a/src/renderer/src/components/editor/rich-markdown/links/rich-markdown-editor-click-routing.ts +++ b/src/renderer/src/components/editor/rich-markdown/links/rich-markdown-editor-click-routing.ts @@ -28,6 +28,7 @@ import type { RichMarkdownHtmlSuperscriptLinkContext } from '../citations/rich-m import type { MutableRefObject } from 'react' import type { Editor } from '@tiptap/react' import type { EditorView } from '@tiptap/pm/view' +import type { Node as ProseMirrorNode } from '@tiptap/pm/model' import type { HttpLinkSourceOwner } from '@/lib/browser/navigation/http-link-routing' import type { DiffComment } from '../../../../../../shared/diff-comment-types' @@ -98,7 +99,7 @@ export function handleRichMarkdownEditorClick({ } return false } - const clickedNode = view.state.doc.nodeAt(pos) + const clickedNode = getClickedNode(view, pos, event) if (clickedNode?.type.name === 'image') { return activateMarkdownImageClick({ activateMarkdownLink, @@ -183,6 +184,19 @@ function activateMarkdownImageClick({ return true } +// posAtCoords resolves a click on an inline atom's right half to the position after it. +function getClickedNode(view: EditorView, pos: number, event: MouseEvent): ProseMirrorNode | null { + const before = view.state.doc.resolve(pos).nodeBefore + if ( + before?.isAtom && + event.target instanceof Node && + view.nodeDOM(pos - before.nodeSize)?.contains(event.target) + ) { + return before + } + return view.state.doc.nodeAt(pos) +} + function getClickedLinkHref(view: EditorView, event: MouseEvent): string { const target = event.target instanceof Element From 134a7be89a00649e9fefe4eb87d69c1d48182ab1 Mon Sep 17 00:00:00 2001 From: Kelvin Amoaba <97001695+AmoabaKelvin@users.noreply.github.com> Date: Thu, 12 Nov 2026 03:11:53 +0000 Subject: [PATCH 08/13] fix(computer): stop macOS type-text and paste-text entering text twice (#41339) * fix(computer): stop macOS type-text and paste-text entering text twice Chromium answers an AX value read from a cache that trails a successful write by about 100 ms. type-text and paste-text read the field once right after writing, saw the old value, and typed or pasted the same text again as keys. set-value read back the same stale value and reported that a landed write did not take effect. Both now wait up to 500 ms for the readback to settle, and text input falls back to keys only when the field still holds its old value. Text input into Chromium web content skips the AX write and sends real keys, because the write fires no beforeinput/input events (none at all for a contenteditable), so editors such as Discord's search never saw the text. * fix(computer): bound the macOS value readback wait by wall clock * docs(computer): say why text readback polls for async AX trees --- .../Sources/OrcaComputerUseMacOS/main.swift | 33 ++++++++++++++++--- .../BackgroundInputTransport.swift | 2 +- 2 files changed, 30 insertions(+), 5 deletions(-) diff --git a/native/computer-use-macos/Sources/OrcaComputerUseMacOS/main.swift b/native/computer-use-macos/Sources/OrcaComputerUseMacOS/main.swift index 082516b941e..f25b5e8b7c9 100644 --- a/native/computer-use-macos/Sources/OrcaComputerUseMacOS/main.swift +++ b/native/computer-use-macos/Sources/OrcaComputerUseMacOS/main.swift @@ -894,7 +894,11 @@ final class Provider { throw ProviderError.coded("accessibility_error", "AXUIElementSetAttributeValue failed with \(result.rawValue)") } let verification: [String: Any] - switch coercion.compare(readback: rawAttributeValue(record.element, kAXValueAttribute as String)) { + let readback = settledValueReadback(record.element) { readback in + if case .mismatch = coercion.compare(readback: readback) { return false } + return true + } + switch coercion.compare(readback: readback) { case let .match(actualPreview): verification = verifiedAction(property: "value", expected: expected, actualPreview: actualPreview) case let .mismatch(actualPreview): @@ -1751,6 +1755,19 @@ private func rawStringAttribute(_ element: AXUIElement, _ attribute: String) -> return value as? String } +/// Reads the value until `settled` accepts it: Chromium answers from a cache that trails a successful write by about 100 ms. +private func settledValueReadback(_ element: AXUIElement, settled: (CFTypeRef) -> Bool) -> CFTypeRef? { + var readback: CFTypeRef? + // Why: settle counts only its sleeps, and one slow or failed AX read can take the full 3 s timeout. + let deadline = CFAbsoluteTimeGetCurrent() + 0.5 + _ = PermissionTrustSettling.settle(timeoutMs: 500, intervalMs: 25) { + readback = rawAttributeValue(element, kAXValueAttribute as String) + guard let readback, CFAbsoluteTimeGetCurrent() < deadline else { return true } + return settled(readback) + } + return readback +} + private func rawAttributeValue(_ element: AXUIElement, _ attribute: String) -> CFTypeRef? { var value: CFTypeRef? guard AXUIElementCopyAttributeValue(element, attribute as CFString, &value) == .success else { @@ -2937,7 +2954,9 @@ private enum Input { private enum TextInput { static func replaceSelection(_ element: AXUIElement, with text: String) throws -> [String: Any]? { - guard isSettable(element, kAXValueAttribute as String), + // Why: Chromium applies AX value writes without the input events pages listen to; a contenteditable gets none, so editors like Discord's never see the text. + guard !BackgroundInputTransport.isChromiumNode(element), + isSettable(element, kAXValueAttribute as String), let current = rawStringAttribute(element, kAXValueAttribute as String) else { return nil @@ -2958,8 +2977,14 @@ private enum TextInput { return nil } setSelectedTextRange(element, CFRange(location: startOffset + text.utf16.count, length: 0)) - guard rawStringAttribute(element, kAXValueAttribute as String) == next else { - return nil + // Why: a cached, out-of-process AX tree can show the old value for a moment; any other value is final. + let readback = settledValueReadback(element) { next == current || ($0 as? String) != current } as? String + guard readback == next else { + // Why: typing the text as keys after a write that landed enters it twice, so only an unchanged field falls back. + if readback == current { + return nil + } + return unverifiedAction(reason: "accessibility_action_unasserted", expected: text, actualPreview: readback.map { preview($0) }) } return verifiedAction( property: "focusedText", diff --git a/native/computer-use-macos/Sources/OrcaComputerUseMacOSCore/BackgroundInputTransport.swift b/native/computer-use-macos/Sources/OrcaComputerUseMacOSCore/BackgroundInputTransport.swift index c66fe0402ca..4c28f717234 100644 --- a/native/computer-use-macos/Sources/OrcaComputerUseMacOSCore/BackgroundInputTransport.swift +++ b/native/computer-use-macos/Sources/OrcaComputerUseMacOSCore/BackgroundInputTransport.swift @@ -326,7 +326,7 @@ public enum BackgroundInputTransport { } } - private static func isChromiumNode(_ element: AXUIElement) -> Bool { + public static func isChromiumNode(_ element: AXUIElement) -> Bool { var names: CFArray? guard AXUIElementCopyAttributeNames(element, &names) == .success, let names = names as? [String] else { return false From 437d078e186f000806594ee966212a034f68515f Mon Sep 17 00:00:00 2001 From: Wooseong Kim <2222333+innocarpe@users.noreply.github.com> Date: Thu, 12 Nov 2026 12:16:46 +0900 Subject: [PATCH 09/13] fix(runtime): close a paired session tab when the desktop window is gone (#41159) * fix(runtime): close a paired session tab when the desktop window is gone Paired session-tab closes carry the device id as a navigation id, and that path captured a renderer graph epoch before the headless retirement. A closed window leaves the graph unavailable, so the close threw runtime_unavailable and the phone showed nothing. Skip the epoch only while the graph is unavailable. A reloading renderer still fails closed. * refactor(runtime): tighten the windowless paired-close fence * fix(runtime): keep a booting desktop renderer fenced for paired tab closes --------- Co-authored-by: Kelvin Amoaba <97001695+AmoabaKelvin@users.noreply.github.com> --- .../mobile-session-tabs-part-05.spec.ts | 85 +++++++++++++++++++ .../session-graph/session-graph-authority.ts | 14 +++ .../close/close-mobile-session-tab.ts | 6 +- ...ersisted-serve-or-ssh-owned-pty-binding.ts | 8 ++ 4 files changed, 110 insertions(+), 3 deletions(-) diff --git a/src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-05.spec.ts b/src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-05.spec.ts index 5a8fd703845..112a2fe60e2 100644 --- a/src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-05.spec.ts +++ b/src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-05.spec.ts @@ -435,4 +435,89 @@ describe('OrcaRuntimeService', () => { expect(getSession().terminalLayoutsByTabId['durable-tab']).toBeUndefined() expect(flushOrThrow).toHaveBeenCalledTimes(1) }) + + // Main-owned layouts restore a ready headless graph when the window closes. + it.skipIf(ORCA_LAYOUT_OWNER_IS_MAIN)( + 'closes a paired session tab after the desktop window is gone', + async () => { + const { runtimeStore, getSession } = makeRuntimeStoreWithWorkspaceSession( + makeWorkspaceSessionWithHeadlessTerminal() + ) + const closeTerminalTab = vi.fn(async () => undefined) + const kill = vi.fn(() => true) + const runtime = new OrcaRuntimeService( + withDurableRuntimeStore({ ...runtimeStore, flushOrThrow: vi.fn() }) + ) + runtime.setNotifier({ closeTerminal: vi.fn(), closeTerminalTab } as never) + runtime.setPtyController({ + write: () => true, + kill, + getForegroundProcess: async () => null, + listProcesses: async () => [] + }) + runtime.syncWindowGraph(1, { + tabs: [ + { + tabId: 'host-tab', + worktreeId: TEST_WORKTREE_ID, + title: 'Durable', + activeLeafId: HEADLESS_LEAF_ID, + layout: null + } + ], + leaves: [ + { + tabId: 'host-tab', + worktreeId: TEST_WORKTREE_ID, + leafId: HEADLESS_LEAF_ID, + paneRuntimeId: 1, + ptyId: 'persisted-pty' + } + ], + mobileSessionTabs: [ + { + worktree: TEST_WORKTREE_ID, + publicationEpoch: 'renderer:paired-close', + snapshotVersion: 1, + activeGroupId: 'group-1', + activeTabId: `host-tab::${HEADLESS_LEAF_ID}`, + activeTabType: 'terminal', + tabs: [ + { + type: 'terminal', + id: `host-tab::${HEADLESS_LEAF_ID}`, + parentTabId: 'host-tab', + leafId: HEADLESS_LEAF_ID, + ptyId: 'persisted-pty', + title: 'Durable', + isActive: true + } + ] + } + ] + }) + runtime.registerPty('persisted-pty', TEST_WORKTREE_ID, null, { + tabId: 'host-tab', + leafId: HEADLESS_LEAF_ID + }) + + const closePaired = () => + runtime.closeMobileSessionTab(`id:${TEST_WORKTREE_ID}`, 'host-tab', { + reason: 'user', + clientNavigationId: 'device-a' + }) + runtime.markGraphUnavailable(1) + runtime.setNotifier(null) + runtime.attachWindow(2) + + await expect(closePaired()).rejects.toThrow('runtime_unavailable') + + runtime.markGraphUnavailable(2) + + await expect(closePaired()).resolves.toEqual({ closed: true, retirementConfirmed: true }) + expect(closeTerminalTab).not.toHaveBeenCalled() + expect(getSession().tabsByWorktree[TEST_WORKTREE_ID] ?? []).toEqual([]) + expect(getSession().terminalLayoutsByTabId['host-tab']).toBeUndefined() + } + ) }) diff --git a/src/main/runtime/session-graph/session-graph-authority.ts b/src/main/runtime/session-graph/session-graph-authority.ts index 13b1e4c30d0..ea078c04082 100644 --- a/src/main/runtime/session-graph/session-graph-authority.ts +++ b/src/main/runtime/session-graph/session-graph-authority.ts @@ -123,6 +123,20 @@ export abstract class OrcaRuntimeWithTransitionGraphReloadToTerminalState extend } } + /** Null when no window is attached: there is no renderer to fence. A booting or reloading one throws. */ + protected captureRendererGraphFence(): number | null { + return this.authoritativeWindowId === null ? null : this.captureReadyGraphEpoch() + } + + protected assertRendererGraphFence(fence: number | null): void { + if (fence !== null) { + this.assertStableReadyGraph(fence) + } else if (this.authoritativeWindowId !== null) { + // A window attached mid-operation; its renderer owns the graph again. + throw new Error('runtime_unavailable') + } + } + protected resolveFolderWorkspaceConnectionId(workspace: FolderWorkspace): string | null { const repos = this.store?.getRepos() ?? [] const projectGroups = this.store?.getProjectGroups?.() ?? [] diff --git a/src/main/runtime/session-lifecycle/close/close-mobile-session-tab.ts b/src/main/runtime/session-lifecycle/close/close-mobile-session-tab.ts index 3b0b7693faa..9b5123c115e 100644 --- a/src/main/runtime/session-lifecycle/close/close-mobile-session-tab.ts +++ b/src/main/runtime/session-lifecycle/close/close-mobile-session-tab.ts @@ -47,14 +47,14 @@ export abstract class OrcaRuntimeWithCloseMobileSessionTab extends OrcaRuntimeWi tabId: string, options: MobileSessionTabCloseOptions = {} ): Promise { - const graphEpoch = options.clientNavigationId ? this.captureReadyGraphEpoch() : null + const graphFence = options.clientNavigationId ? this.captureRendererGraphFence() : undefined const explicitWorktreeId = this.getExplicitSessionWorktreeIdSelector(worktreeSelector) const worktreeId = explicitWorktreeId ?? (await this.resolveWorkspaceSelector(worktreeSelector)).id this.hydrateHeadlessMobileSessionTabsFromWorkspaceSession(worktreeId) const observedPtyIds = await this.refreshMobileSessionPtyRecords() - if (graphEpoch !== null) { - this.assertStableReadyGraph(graphEpoch) + if (graphFence !== undefined) { + this.assertRendererGraphFence(graphFence) } this.restoreLivePairedRendererSessionOwnedMobileTerminals(worktreeId) const snapshot = this.mobileSessionTabsByWorktree.get(worktreeId) diff --git a/src/main/runtime/session-lifecycle/close/has-live-or-persisted-serve-or-ssh-owned-pty-binding.ts b/src/main/runtime/session-lifecycle/close/has-live-or-persisted-serve-or-ssh-owned-pty-binding.ts index 90d665e1af0..3475a48d110 100644 --- a/src/main/runtime/session-lifecycle/close/has-live-or-persisted-serve-or-ssh-owned-pty-binding.ts +++ b/src/main/runtime/session-lifecycle/close/has-live-or-persisted-serve-or-ssh-owned-pty-binding.ts @@ -26,6 +26,14 @@ export abstract class OrcaRuntimeWithHasLiveOrPersistedServeOrSshOwnedPtyBinding ...args: Parameters ): ReturnType + protected abstract captureRendererGraphFence( + ...args: Parameters + ): ReturnType + + protected abstract assertRendererGraphFence( + ...args: Parameters + ): ReturnType + protected abstract resolvePtyTabCloseSurfaceAuthority( ...args: Parameters ): ReturnType From 333febd77bbc67ae1a29d0177a96e6e2009846e0 Mon Sep 17 00:00:00 2001 From: OrcaWin Date: Wed, 11 Nov 2026 19:22:20 -0800 Subject: [PATCH 10/13] test: remove narrow mobile, provider, and utility suites (#41342) * test: remove low-value mobile classification and wiring cases * test: remove narrow provider identity and utility suites --------- Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com> --- config/reliability-gates.jsonc | 25 +-- mobile/src/platform/host-os.web.test.ts | 38 ---- ...leNativeChatComposer.keyboard.web.test.tsx | 163 ------------------ ...erver-kiro-program-status-identity.test.ts | 64 ------- .../codex/server-codex-session-start.test.ts | 157 ++++++++--------- .../server-opencode2-program-status.test.ts | 33 +--- .../artifacts/artifact-cloud-config.test.ts | 31 +--- .../claude-conversation-title-owner.test.ts | 23 --- src/shared/app-version.test.ts | 14 -- src/shared/speech-model-suggestion.test.ts | 44 ----- 10 files changed, 91 insertions(+), 501 deletions(-) delete mode 100644 mobile/src/platform/host-os.web.test.ts delete mode 100644 src/main/agent-hooks/server/testing/identity/server-kiro-program-status-identity.test.ts delete mode 100644 src/main/claude/structured/claude-conversation-title-owner.test.ts delete mode 100644 src/shared/app-version.test.ts delete mode 100644 src/shared/speech-model-suggestion.test.ts diff --git a/config/reliability-gates.jsonc b/config/reliability-gates.jsonc index a15b4a546f1..084505ed081 100644 --- a/config/reliability-gates.jsonc +++ b/config/reliability-gates.jsonc @@ -4058,10 +4058,10 @@ "providers": ["remote-runtime"], "coveredPlatforms": ["macos"], "coveredProviders": ["remote-runtime"], - "coverageNotes": "Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3460 final-v2 composes accepted AUTHOR3454/PEER3455 on actual post263 main: intentionally removes99 authored/106 potential expanded cases across16 selected drivers, including7 whole suites, retaining24 authored/26 potential cases. No equivalent replacement or fresh runtime is claimed. Broad root-unit discovery loses whole cross-client workspace-sync, terminal-preview, picker-skill-hint and close-routing suites plus reduced worker-stop, remote-owned-record, scanner, mobile-bundle and daemon-health samples. Separate E2E discovery also loses the paired mobile/desktop mirror journey and two unit drivers. Retained callback/table/outer-loop/guard bytes are exact; retained checks do not replace the removed alarms. No app, native device, mixed-version, remote or cross-platform execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit1640 removes low-value IPC, SSH, workspace and transcript-watcher tests without equivalent replacement coverage. Per-source reviews record retained public refusal, cancellation, execution-host and real-filesystem outcomes with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Deterministic main-IPC contract tests cover disconnect-driven close delivery, exactly-once close, per-subscription teardown isolation against a failing socket close and a throwing liveness probe, containment of a throwing renderer send on the unguarded host-close path, and continued suppression of stale payloads from a retired transport. A headed paired-server journey (real Orca host plus a separate paired Orca desktop client) covers hidden-but-mounted reveal, cold-parked reveal, and cold-parked reveal across a disconnect/reconnect. Live Linux and Windows paired-server evidence and real sleep/wake transport loss remain uncollected.", + "coverageNotes": "Audit3563 scope: retire the whole Kiro OSC/hook identity suite and the OpenCode2 identity-only case; reduce Codex cached startup child-retirement and missing-session-ID tables to one case each (2 authored/4 finite removals). Intentionally lose Kiro alias/session continuity, OpenCode2 completed OSC identity continuity, cached replacement without relay-generation restart, and missing-previous-ID context-preservation alarms. Remaining provider, remote compatibility, persistence and cross-session pollution cases are not equivalent replacements. Also remove the private Codex roster-size assertion while retaining public child-absence checks. Historical evidence retains its original source scope; no new execution or transferred pass validates these reductions. Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3460 final-v2 composes accepted AUTHOR3454/PEER3455 on actual post263 main: intentionally removes99 authored/106 potential expanded cases across16 selected drivers, including7 whole suites, retaining24 authored/26 potential cases. No equivalent replacement or fresh runtime is claimed. Broad root-unit discovery loses whole cross-client workspace-sync, terminal-preview, picker-skill-hint and close-routing suites plus reduced worker-stop, remote-owned-record, scanner, mobile-bundle and daemon-health samples. Separate E2E discovery also loses the paired mobile/desktop mirror journey and two unit drivers. Retained callback/table/outer-loop/guard bytes are exact; retained checks do not replace the removed alarms. No app, native device, mixed-version, remote or cross-platform execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit1640 removes low-value IPC, SSH, workspace and transcript-watcher tests without equivalent replacement coverage. Per-source reviews record retained public refusal, cancellation, execution-host and real-filesystem outcomes with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Deterministic main-IPC contract tests cover disconnect-driven close delivery, exactly-once close, per-subscription teardown isolation against a failing socket close and a throwing liveness probe, containment of a throwing renderer send on the unguarded host-close path, and continued suppression of stale payloads from a retired transport. A headed paired-server journey (real Orca host plus a separate paired Orca desktop client) covers hidden-but-mounted reveal, cold-parked reveal, and cold-parked reveal across a disconnect/reconnect. Live Linux and Windows paired-server evidence and real sleep/wake transport loss remain uncollected.", "motivatingLinks": ["tests/e2e/paired-remote-terminal-parked-reveal-interactivity.spec.ts"], "invariant": "Every renderer-held runtime subscription receives exactly one terminal close event when its transport is retired, including when the retirement advanced the transport generation first, and a single failing teardown never abandons that environment's remaining subscriptions nor escapes into the transport that reported the close. Payload frames from a retired transport stay suppressed. A revealed remote terminal therefore reattaches over a live multiplex connection: its buffer restores, typed input reaches the host PTY, the echo paints without a tab flip, and the PTY converges on the revealed pane grid.", - "oracle": "Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3460 final-v2 composes accepted AUTHOR3454/PEER3455 on actual post263 main: intentionally removes99 authored/106 potential expanded cases across16 selected drivers, including7 whole suites, retaining24 authored/26 potential cases. No equivalent replacement or fresh runtime is claimed. Broad root-unit discovery loses whole cross-client workspace-sync, terminal-preview, picker-skill-hint and close-routing suites plus reduced worker-stop, remote-owned-record, scanner, mobile-bundle and daemon-health samples. Separate E2E discovery also loses the paired mobile/desktop mirror journey and two unit drivers. Retained callback/table/outer-loop/guard bytes are exact; retained checks do not replace the removed alarms. No app, native device, mixed-version, remote or cross-platform execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): The main IPC contract test subscribes terminal.multiplex through the real handler, disconnects the environment, and asserts the renderer received exactly one {type: close} subscription event. Two isolation tests subscribe a second stream to the same environment and make the first one fail -- in its socket close, and in the liveness probe inside notifyClosed -- then assert the disconnect does not throw, both transports closed, and every close the renderer could still receive was delivered. A third drives a host-initiated close through the transport callback, which is the one notifyClosed call site with no surrounding guard, with a renderer send that throws, and asserts it cannot escape into the WebSocket close handler. A fourth test asserts that after retirement a late response frame is not forwarded and a late transport close does not re-send. The paired-server journey runs three reveal scenarios against one real host and one real paired desktop client, and for each records buffer restore, host-side receipt of the typed marker through an out-of-band host sink file, live paint without a tab flip, and PTY-versus-pane grid convergence.", + "oracle": "Audit3563 scope: retire the whole Kiro OSC/hook identity suite and the OpenCode2 identity-only case; reduce Codex cached startup child-retirement and missing-session-ID tables to one case each (2 authored/4 finite removals). Intentionally lose Kiro alias/session continuity, OpenCode2 completed OSC identity continuity, cached replacement without relay-generation restart, and missing-previous-ID context-preservation alarms. Remaining provider, remote compatibility, persistence and cross-session pollution cases are not equivalent replacements. Also remove the private Codex roster-size assertion while retaining public child-absence checks. Historical evidence retains its original source scope; no new execution or transferred pass validates these reductions. Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3460 final-v2 composes accepted AUTHOR3454/PEER3455 on actual post263 main: intentionally removes99 authored/106 potential expanded cases across16 selected drivers, including7 whole suites, retaining24 authored/26 potential cases. No equivalent replacement or fresh runtime is claimed. Broad root-unit discovery loses whole cross-client workspace-sync, terminal-preview, picker-skill-hint and close-routing suites plus reduced worker-stop, remote-owned-record, scanner, mobile-bundle and daemon-health samples. Separate E2E discovery also loses the paired mobile/desktop mirror journey and two unit drivers. Retained callback/table/outer-loop/guard bytes are exact; retained checks do not replace the removed alarms. No app, native device, mixed-version, remote or cross-platform execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): The main IPC contract test subscribes terminal.multiplex through the real handler, disconnects the environment, and asserts the renderer received exactly one {type: close} subscription event. Two isolation tests subscribe a second stream to the same environment and make the first one fail -- in its socket close, and in the liveness probe inside notifyClosed -- then assert the disconnect does not throw, both transports closed, and every close the renderer could still receive was delivered. A third drives a host-initiated close through the transport callback, which is the one notifyClosed call site with no surrounding guard, with a renderer send that throws, and asserts it cannot escape into the WebSocket close handler. A fourth test asserts that after retirement a late response frame is not forwarded and a late transport close does not re-send. The paired-server journey runs three reveal scenarios against one real host and one real paired desktop client, and for each records buffer restore, host-side receipt of the typed marker through an out-of-band host sink file, live paint without a tab flip, and PTY-versus-pane grid convergence.", "commands": [ "pnpm exec vitest run --config config/vitest.config.ts", "pnpm exec playwright test tests/e2e/paired-remote-terminal-parked-reveal-interactivity.spec.ts --config tests/playwright.config.ts --project electron-headless --workers=1" @@ -4129,6 +4129,7 @@ "Add real sleep/wake and network-loss transport drops alongside the explicit disconnect trigger." ], "knownGaps": [ + "Audit3563 scope: retire the whole Kiro OSC/hook identity suite and the OpenCode2 identity-only case; reduce Codex cached startup child-retirement and missing-session-ID tables to one case each (2 authored/4 finite removals). Intentionally lose Kiro alias/session continuity, OpenCode2 completed OSC identity continuity, cached replacement without relay-generation restart, and missing-previous-ID context-preservation alarms. Remaining provider, remote compatibility, persistence and cross-session pollution cases are not equivalent replacements. Also remove the private Codex roster-size assertion while retaining public child-absence checks. Historical evidence retains its original source scope; no new execution or transferred pass validates these reductions.", "Launch direct-owner cleanup current scope: This gate’s unfiltered unit command now selects a suite reduced by 82 whole registrations / 159 potential table examples in 10 whole test Programs, including 8 registrations / 15 examples from five once-registered create-message factory closures. Lost alarms include frozen first-message/options and legacy-host fallback; distinct actions and redelivery; uncertain/refused/reloaded input custody, Retry and late proof; rich prompt/attachment upload Stop/close and handback; ordinary draft/editor/IME/StrictMode lifetime and accepted topology; provisional fallback tab placement/group/pin/order and rich worktree handoff. Four whole consent/resume-identity cases remain in two Programs; they do not replace these alarms. This entry governs current reduced-unit scope and qualifies any broader implication of unchanged oracle/coverageNotes history. Entire dated evidenceRuns, original commands/currentCommand values, counts/results and existing history retain their original scopes only; no fresh execution, transferred pass, replacement equivalence or prior upstream-removal credit is claimed.", "Audit2133 current-parent preparation (988 authored removals; independently reviewed2099/2107/2115/2116 decisions; upstream clear-buffer and recovery-image regressions retained): Intentionally retire these low-value alarms without replacement: src/main/persistence/worktree-sort-order-write.test.ts: re-ranks every writable row when a blocked row sits in the requested order. src/main/runtime/construction/orca-runtime.test.ts: remove only literal registration import/call pairs await import('../orca-runtime-tests/worktree-setup-and-startup.spec'), await import('../orca-runtime-tests/gitlab-and-pr-bases.spec'); no additional authored credit. src/main/runtime/orca-runtime-tests/targeting-and-resilience.spec.ts: routes tab switch and capture start by explicit page id; accepts focus on tab switch without altering bridge args (focus is main-side concern); keeps terminal cwd resolution fail-soft when the provider is unavailable. src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-04.spec.ts: drops the bounded scan cache when orphan-cleanup removal completes; clears optimistic reconcile tokens when a CLI worktree removal succeeds; emits drift without mutating when the fetched base fast-forwards created HEAD; emits base_changed without mutation when the fetched base rewrote history; skips stale-token reconciles without mutating or emitting stale status; emits unknown without mutation when fetch fails or the base ref is missing. src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-03.spec.ts: routes runtime preserved-branch force-delete through the selected WSL project runtime; replies to a waiting client with the error of a delete that fails after acceptance; treats forced runtime deletion of an already-missing unregistered worktree as cleanup; treats normal runtime deletion of an already-missing unregistered worktree as cleanup. src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-02.spec.ts: warns that a missing-repo removal only forgot the workspace; restores runtime watchers when CLI worktree deletion fails after teardown; does not re-spend the drain budget on a removal after a wedged install was abandoned; routes runtime worktree removal through the selected WSL project runtime; force-deletes a branch that was preserved by runtime worktree removal. src/main/runtime/orca-runtime-tests/gitlab-and-pr-bases-part-02.spec.ts: records GitLab pasted-project recents only after successful runtime lookup; threads explicit origin preference through runtime WSL PR base resolution; resolves local GitLab fork MR bases from the target project MR head ref; captures the fork MR head from a dedicated ref, not the shared FETCH_HEAD; keeps the durable MR head when the head fetch fails but the local ref resolves; routes runtime GitLab fork MR base git calls through the selected WSL project runtime. src/main/runtime/orca-runtime-tests/gitlab-and-pr-bases.spec.ts: passes SSH connection ids through GitLab task operations; routes runtime GitLab issue, MR, work-item, and todo actions through the selected WSL project runtime; routes runtime GitLab MR details, review-management, job, and pasted URL actions through the selected WSL project runtime; normalizes runtime GitLab issue list arguments like the desktop IPC path. src/main/runtime/orca-runtime-tests/worktree-setup-and-startup-part-03.spec.ts: does not warn when setup is explicitly skipped for CLI-created worktrees; puts the agent a create starts in the first default tab, as the window does; keeps the 8s main-runtime startup readiness budget for agents without an override; gives OpenCode the 20s main-runtime startup readiness budget; launches explicit startup agents with prompts for CLI-created worktrees. src/main/runtime/orca-runtime-tests/worktree-setup-and-startup.spec.ts: returns a setup launch payload for CLI-created worktrees when hooks are explicitly enabled; does not hand the window a setup main never tried to start; passes the selected Windows setup shell into runtime runner generation; follows normal setup policy for CLI-created worktrees without activating them; uses returned WSL setup shell metadata when runtime spawns setup; uses the shell-aware setup runner for windowless creates without a startup command; reports the in-process setup hook as running when nothing can launch the runner. src/main/runtime/orca-runtime-tests/terminal-sleep-and-teardown-part-02.spec.ts: coalesces two clients sleeping the same host worktree; stops exactly the expected live PTYs for a worktree; explains a refused workspace close instead of throwing the raw refusal enum. src/main/runtime/orca-runtime-tests/mobile-summaries-part-03.spec.ts: marks the desktop-active worktree as isActive; includes SSH-backed worktrees in the mobile worktree summary; caches repeated malformed path misses before normalizing them again. src/main/runtime/orca-runtime-tests/mobile-summaries-part-02.spec.ts: keeps title-only foreground work ahead of monitoring in another pane; keeps a local hook row while a connected PTY still backs its pane; keeps an OSC row via its connected PTY after the pane binding is cleared. src/main/runtime/orca-runtime-tests/mobile-summaries.spec.ts: resolves files through a source-visible linked-checkout worktree; marks saved session tabs with live PTYs as host sidebar activity; attributes live legacy PTYs from saved layout bindings when their panes are hidden; marks saved browser tabs as host sidebar activity like desktop; falls back to the path-keyed GitHub cache entry; includes folder workspaces in compact worktree summaries for mobile; attaches inline agent rows from the latest OSC 9999 status; projects monitoring over a title-derived working status. src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-03.spec.ts: pushes to an idle pane when the only live waiter filters out the message type; resolves a registered waiter without PTY-injecting when the leaf is already idle; keeps already-idle status after tui-idle wait for immediate message delivery; resolves message waiters when notifyMessageArrived is called; does not resolve type-filtered message waiters for unrelated message types; removes message waiter abort listeners after message arrival; resolves message waiters on timeout when no message arrives. src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-02.spec.ts: delivers pending mail via notifyMessageArrived when the recipient is already idle; does not keep retrying when every pending row was already pointed; repoints pending rows restored with a live-idle coordinator. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-13.spec.ts: rolls back a half-created terminal whose surface never publishes. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-12.spec.ts: rejects startup prompts for agents that require post-ready stdin; uses portable Unix quoting for mobile agent launch commands in WSL project runtimes; keeps PowerShell quoting for mobile agent launch commands in Windows host runtimes; uses cmd.exe quoting for mobile agent launch commands in local Windows host runtimes; publishes headless mobile session agent identity with synthesized PTY status; rejects disabled mobile session agent launches before spawning; validates mobile terminal insertion anchors before resolving agent launch commands. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-11.spec.ts: coalesces a reconnect close burst onto one authoritative PTY inventory; keeps an explicit user close destructive while the PTY is live; builds mobile session agent launch commands on the runtime host; injects mobile quick-command prompts into the host-built agent startup command. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-10.spec.ts: closes persisted headless terminal parents before any prior list call; tears down a serve-owned headless tab on close while a renderer is attached so it cannot resurrect; defers a renderer-published pending tab to the renderer instead of tearing it down; returns a delegated close outcome with no selection tombstone when the renderer owns the outcome; tears down a runtime pending shell the renderer never adopted on close. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-08.spec.ts: materializes a plain shell when the pending tab has no launch agent; hydrates legacy persisted terminal tabs without layout entries; reads and indexes the full agent-status snapshot once per mobile projection; kills persisted SSH PTYs when closing hydrated headless tabs before pane metadata is restored; materializes hydrated pending headless terminals with the persisted session identity. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-07.spec.ts: hydrates a persisted SSH-owned pane before an attached renderer publishes its graph. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-05.spec.ts: publishes the desktop's tab name over the live agent title, and the live title without one; uses fresh neutral PTY titles over stale mobile snapshot and OSC titles; operates PTY-backed mobile session terminals without a renderer graph; reuses pane close for live PTYs that do not own a renderer tab; lists PTY-backed mobile session terminals without a renderer graph; shows and resolves active PTY-backed mobile session terminals without a renderer graph. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-04.spec.ts: creates mobile session terminals for folder workspaces in a headless runtime server; spawns fresh headless SSH mobile session terminals instead of reattaching synthetic local ids; hydrates headless mobile session terminals from the host workspace session; keeps live headless mobile session terminals when a desktop renderer publishes without them; publishes laptop-created remote runtime split terminals to phone session tabs. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-02.spec.ts: does not report headless wake degradation without sleeping records; does not wake slept agents for non-mobile session-only activation; does not rewrite unread metadata when a mobile activation finds the worktree already read; returns unread:false from worktree.ps after a mobile activation clears the flag; materializes pending mobile session terminals without focusing desktop clients; closes browser mobile session tabs when addressed by browser workspace id. src/main/runtime/orca-runtime-tests/mobile-session-tabs.spec.ts: resolves mobile terminal surfaces by exact split leaf; closes the matching mobile terminal UUID leaf without closing the whole tab; closes the whole mobile terminal tab when addressed by parent tab id; activates the active split leaf when addressed by parent tab id; clears unread metadata on mobile worktree activation without focusing desktop clients; wakes slept agents on the host renderer when a phone activates a worktree. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-05.spec.ts: publishes the hook provider session on a headless mobile tab so native chat can address the transcript; recovers the agent type from the hook row when the pane was launched without an agent hint; reads one agent-status snapshot per projection, not one per terminal tab; waits for unknown-launch foreground owner before publishing Pi-compatible mobile status; keeps Pi-compatible mobile session status as Pi for an unknown-launch foreground pi PTY; keeps renderer-vetted mobile agent status for custom-titled terminals. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-04.spec.ts: publishes terminal surface agent status for paired web clients; preserves authoritative OMP identity for Pi-compatible remote terminal snapshots; normalizes a remote OMP title without republishing omitted launch identity; skips the foreground-process probe when the PTY launch agent is already known; probes the foreground process only on a status transition for unknown launch agents; passes the client signal and the slept-worktree refusal to the spawn. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-03.spec.ts: does not use stale runtime-created PTY status when a neutral PTY title exists; recognizes runtime-created Codex PTY handles from the ready prompt; recognizes runtime-created Antigravity PTY handles from the ready prompt; recognizes live leaf Antigravity terminals from the ready prompt; recognizes a later Antigravity header with a prompt but no model line; keeps mobile terminal surfaces visible while their leaf handle is pending. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-02.spec.ts: lets Claude agents management titles clear stale runtime-created title status; does not recognize live Claude agents panes from a Claude foreground process; does not let stale tab-level agent titles override current neutral pane titles; lists live terminals with fresh pane titles over stale tab titles; lets adopted neutral pane titles use non-shell foreground fallback; lets adopted neutral pane titles retry wrapper foregrounds until recognized; authorizes settled CLI prompts only after positive %s foreground identity; does not recognize arbitrary foreground TUIs as running agents; lets live neutral pane titles retry wrapper foregrounds until recognized; lets current Claude agents PTY titles override stale runtime-created OSC titles; does not let stale Claude agents OSC titles suppress current PTY title activity. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status.spec.ts: keeps preallocated terminal handles valid across renderer reloads; preserves runtime-created PTY process identity after graph unavailable; keeps prompt bindings fenced across runtime restarts without provider incarnation; recognizes runtime-created PTY handles with agent launch titles; does not treat a bare Cursor Agent native title as a running agent session; uses stale runtime-created PTY status when there is no title or foreground evidence. src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-07.spec.ts: keeps orphaned list and show writability aligned with the send gate. src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-04.spec.ts: cancels a coalesced SSH worker recovery retry when its provider disconnects. src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-03.spec.ts: retries renderer reveal before clearing an adopted legacy worker sleeping record. src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-02.spec.ts: recovers exported ORCA_TERMINAL_HANDLE from discovered live PTY sessions. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-12.spec.ts: applies ANSI erase-from-start line controls in retained previews; applies ANSI stripping for private or intermediate CSI line controls; applies ANSI stripping for unsupported erase-line modes; reveals an agent terminal under the pane the caller reserved; keeps a fresh %s pane host-owned when a desktop window is available; keeps desktop publication %s separate from the reserved host spawn; mints its own pane when the reserved leaf is malformed, and reports the one it used; starts an agent terminal with the model the caller picked; starts an agent terminal without a model flag when no pick was made; attributes a startup-agent launch named %s as %s; attributes a fresh agent session the host builds as unknown; leaves a bare agent command the user typed out of launch attribution. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-11.spec.ts: trims oversized terminal output bursts without per-line array shifts; builds terminal previews without mapping the full retained tail; keeps recent PTY replay output capped without needing previous data for large chunks; applies terminal redraw controls before retaining previews; applies ANSI terminal redraw controls before retaining previews; retains multi-line ANSI footer redraws instead of appending old frames; keeps the cursor column when erasing full lines in multi-line redraws; applies ANSI split redraw controls without leaking raw params. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-10.spec.ts: returns renderer visible screen lines through terminal.read RPC JSON result; separates composer drafts from provider-owned terminal screens. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-09.spec.ts: reports an in-flight PTY focus as not navigated when its notifier disappears; reports graph-leaf focus as not navigated without a host notifier; grounds the provider mode tracker before a later chunk, in the emulator order; grounds an in-flight provider snapshot capture so it cannot publish the pre-reset screen; waits for terminal exit and resolves with the exit status; keeps terminal read payloads bounded while retained output remains pageable; trims terminal read preview character budget without per-line array shifts. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-08.spec.ts: submits a silent Claude composer once after the bounded render fallback; chunks large terminal.send text before provider writes; yields chunked terminal input through immediates between writes; yields while validating accepted large terminal.send text before provider writes; reveals a background terminal session when focusing its handle; replays captured launch config when focusing a background agent session; reveals background terminal sessions with the freshest PTY title; renames background terminal handles without requiring a visible tab; keeps a background terminal handle stable while reveal adoption is racing. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-07.spec.ts: resolves tui-idle when a stale Codex prompt is followed by Antigravity readiness; returns an agent-neutral blocked wait result for update prompts; returns an agent-neutral blocked wait result for workspace trust prompts; returns an agent-neutral blocked wait result for cwd selection prompts; returns a blocked wait result for Codex model migration prompts; returns a blocked wait result for Codex startup hook review prompts; returns an agent-neutral blocked wait result for generic interactive prompts; splits text and enter writes for background terminal handles; types the lead line for %s: %s; submits through the agent-specific PTY timing policy for %s. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-06.spec.ts: waits for exit on background terminal handles; keeps retained PTY transcript memory when controller refresh fails; resolves tui-idle from an Antigravity ready prompt preview; resolves Antigravity ready prompts with newline-heavy pasted tails without splitting; resolves live-leaf tui-idle from an Antigravity ready prompt preview. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-03.spec.ts: spawns the available Qoder command for a captured same-session resume: %s; does not use the local Windows shell setting for remote Windows bare agent creates; matches canonical bare agent commands when a command override is configured; keeps non-bare agent command terminal creates unchanged; keeps disabled bare agent command terminal creates unchanged; injects runtime hook receiver env into terminal sessions; creates background terminal sessions for a $label; creates background terminal sessions for a $label; enables Claude Agent Teams only for direct Claude launches when configured in-process; reveals Claude Agent Teams launches with the rewritten launch config; preserves Claude Agent Teams for sequenced Claude launches. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness.spec.ts: creates visible terminal sessions without asking the renderer to focus a tab; asks the pty controller for the requested shell instead of a startup command; knows the agent a flagged launch command names before any process read; retires a flagged command pane identity when the command exits; retires only receipted restored PTY authority on command completion and exit; forgets synthetic handles when disconnected PTY records are pruned. src/main/runtime/orca-runtime-tests/headless-snapshots.spec.ts: returns OSC titles from headless main terminal snapshots; adopts renderer-seeded titles into headless main terminal snapshots; returns cwd metadata from seeded headless main terminal snapshots; adopts OSC7 host metadata from seeded headless terminal scrollback; falls back to the renderer snapshot for hidden-output recovery without headless state; keeps an empty headless snapshot authoritative for hidden-output recovery; advances the absolute output sequence across a daemon stream gap; emits explicit OSC 9999 agent status from runtime PTY data. src/main/runtime/orca-runtime-tests/terminal-side-effect-facts-part-02.spec.ts: falls back to the provider snapshot for a restored PTY with no mounted renderer; falls back to the available mirror when authoritative provider history is unavailable; keeps an empty renderer snapshot when the provider has no retained content; keeps restored provider history authoritative after later live output; uses provider alternate-screen state while a partial model is unsafe; prefers the tracked title over the headless emulator lastTitle; keeps the cursor-agent literal in record-fallback snapshots only without a tracker title; touches mobile snapshots once for decorative spinner ticks, again on idle. src/main/runtime/orca-runtime-tests/local-worktree-creation-part-02.spec.ts: creates a same-repo PR branch override from a resolved head SHA and matching push target; skips broad remote fetch for an existing full-SHA PR base; creates a selected Bitbucket PR branch override from a matching remote branch; suffixes an existing PR when a matching push target lacks selected PR metadata; suffixes a matching push target branch when selected PR metadata has no PR number; suffixes a matching push target branch when the existing PR is different; suffixes a selected PR remote conflict when the PR lookup fails; checks out an unused runtime PR branch only when it is at the resolved head SHA; suffixes only the runtime worktree path when an exact PR branch checkout path exists. src/main/runtime/orca-runtime-tests/local-worktree-creation.spec.ts: returns runtime local base update suggestions from addWorktree; creates a runtime local worktree from a usable persisted local branch base; does not create a runtime local worktree when the refresh fails and no local base ref exists; creates a branchNameOverride worktree from the selected matching remote base ref. src/main/runtime/integrations/skill/orca-runtime-skill-recovery.test.ts: continues skill management after startup recovery fails. src/main/runtime/orchestration/mailbox/routing/orchestration-mailbox-detached-routing.spec.ts: uses bounded ownership lookup for direct arrivals while the agent cannot take input; wakes a bound Run waiter when same-Run direct mail arrives; keeps an unowned detached-recipient message in its direct mailbox; uses primary-key lookups for bounded routing updates. src/main/runtime/orchestration/mailbox/terminal-pointer/orchestration-mailbox-delivery.test.ts: remove only literal registration import/call pairs import './orchestration-mailbox-pointer-cli-command.spec'; no additional authored credit. src/main/runtime/orchestration/mailbox/terminal-pointer/orchestration-mailbox-pointer-cli-command.spec.ts: renders the %s CLI command in a mailbox pointer. src/main/runtime/rpc/methods/terminal/terminal-multiplex-link-estimate.test.ts: sizes a slow high-latency link near twice its bandwidth-delay product; never paces a LAN, even when its client parses slowly; never paces a fast link; returns to legacy windows when the path RTT jumps. src/main/runtime/rpc/methods/terminal/terminal-snapshot-scrollback-rows.test.ts: steps down the row ladder when history is over the byte budget; tries the larger first row count, then the usual ladder within the same byte budget. src/main/runtime/rpc/server/pairing/network/windows-default-route-interfaces.test.ts: returns unknown when route inspection fails or emits malformed data; does not inspect routes outside Windows. src/main/runtime/rpc/server/pairing/pairing-offer-owner-cli.test.ts: is refused to a paired %s client; reports a failed widening instead of advertising an unreachable address. src/main/runtime/rpc/terminal-multiplex-end-verdict.test.ts: reports exited only when the owning runtime completes the exit waiter; reports unverifiable when the runtime cannot observe the exit waiter; reports unverifiable when a disconnected PTY resolves with unknown liveness. src/main/runtime/session-graph/headless-runtime-graph.spec.ts: lets a session tab inventory settle on a host no renderer publishes for. src/main/runtime/session-graph/runtime-graph-standing-contracts.test.ts: remove only literal registration import/call pairs import { registerHeadlessRuntimeGraphTests } from './headless-runtime-graph.spec', registerHeadlessRuntimeGraphTests(); no additional authored credit. src/main/runtime/session-tabs/structured-conversation-tab-title.test.ts: bumps the stored snapshot version only when the title changes. src/main/runtime/session-tabs/terminal-close-observed-exit.test.ts: keeps the healthy aggregate verification and delayed physical exit behavior; uses a stamped exit for the incarnation that was actually being closed. src/main/runtime/terminal/identity/terminal-identity-probe.test.ts: answers live for a structured worker without touching the PTY graph; answers live for a PTY handle the runtime still holds; runs the full leaf check for a handle with no live PTY; answers not-live for a stale handle; propagates \"could not look\" rather than reporting it as a dead handle. src/main/runtime/terminal/layout/paired-desktop-server-writes-layout-owner.test.ts: names the tab in main’s layout through terminal.rename; takes a window-style layout.apply close as already stopped, so the terminal runs on. src/main/runtime/terminal/lifecycle/pty-lifecycle-generation-retention.test.ts: retains no lifecycle generation after a spawn/exit cycle. src/main/runtime/terminal/lifecycle/run-terminal-user-open.test.ts: reads a freshly spawned terminal as used once a user opens its tab, not on a probe; keeps the open mark on the process a user open starts. src/main/runtime/terminal/surface/terminal-surface-replies-window.test.ts: from a desktop window that mounts both panes; from a server with no window. src/main/runtime/terminal/surface/workspace-layout-projection.test.ts: carries views for a partition that holds none yet. src/main/runtime/terminal/view/terminal-subscribe-exit-waiter-leak.test.ts: releases a signalled exit-waiter when the signal aborts. src/main/runtime/terminal/wait/orca-runtime-exact-worker-provider-session.test.ts: selects a local hook status for a local pane. src/main/runtime/terminal/wait/prime-agent-screen-readiness-transcripts.test.ts: keeps the footer and the bare caret on screen mid-turn; %s: the screen alone reads ready during %s. src/main/runtime/workspaces/startup/runtime-worktree-agent-startup.test.ts: uses per-launch arguments and preserves launch telemetry; attributes a startup agent whose caller named no surface as unknown; launches %s bare for an absent or blank prompt; attributes a %s draft launch named %s as %s; merges an automation's extras over the host defaults; threads worktree-create extras into the startup build; returns an enabled requested agent without detecting; returns the default when nothing was requested; detects instead of using the default when the requested agent is disabled; starts no agent when the default is blank; detects an enabled agent on this host when nothing usable was named; starts no agent when detection finds none. src/main/runtime/workspaces/startup/runtime-worktree-startup-readiness.test.ts: pastes and submits through composer readiness, never typing raw input; types the follow-up on WSL or SSH-to-Windows once the agent is in front, as before; cleans up the deadline when the composer was already captured; a draft that is not submitted is pasted on the box cursor, as before; takes the box after the grace when the row never comes, whatever repaints; takes the box at the deadline when the grace would outlast the budget; lets a dialog on screen veto a grace; refuses the deadline inside a pending grace while a shell is in front; takes the box at the deadline inside a pending grace with the agent in front; takes the box at the deadline after a refused grace once the agent is in front. src/main/skills/skill-install-transaction.test.ts: installs, previews unchanged bytes, and updates a clean receipt-owned skill; preserves the old version after a %s filesystem failure. src/main/speech/model-manager-download-resume.test.ts: requests the remaining bytes when a clean range response ends before the file total; bounds a server that advances by pathologically tiny segments forever. src/main/ssh/config/ssh-config-parser.test.ts: parses a basic host block; parses multiple host blocks; skips wildcard-only Host entries; skips Host entries with only pattern characters; parses IdentityFile with ~ expansion; parses Windows-style IdentityFile with ~ expansion; parses quoted scalar values like OpenSSH; parses equals-form scalar directives; parses IdentityAgent with ~ expansion; parses IdentitiesOnly; parses GSSAPIAuthentication; keeps the first GSSAPIAuthentication value like OpenSSH; parses ProxyCommand, ProxyUseFdpass, and ProxyJump; ignores comments and blank lines; handles case-insensitive keywords; stops current block on Match directive; returns empty array for empty input; creates one parsed host per concrete alias on a multi-pattern Host line; applies identity agent settings to every concrete alias on a multi-pattern Host line; defaults port to 22 for invalid port values; converts hosts to SshTarget objects; uses host alias as hostname when HostName is missing; skips hosts that are already imported; defaults username to empty string when not specified; carries through identityFile, identityAgent, identitiesOnly, proxyCommand, and jumpHost; carries through gssapiAuthentication; imports duplicate aliases only once and keeps the first concrete host; maps hosts for the picker and flags aliases already in Orca; de-dupes aliases and defaults missing fields; parses hostname, user, port from ssh -G output; collects multiple identity files; parses forwardagent yes; defaults port to 22 when missing; returns empty hostname when missing; returns undefined user when missing; handles empty output; skips lines without spaces; parses proxycommand and filters \"none\"; parses proxyusefdpass yes; parses proxyjump and filters \"none\"; handles ~ expansion in identity file paths; handles Windows-style ~ expansion in identity file paths; parses identityagent with ~ expansion; preserves identityagent none so auth can disable agent fallback; parses identitiesonly yes; parses gssapiauthentication; parses controlmaster options and filters controlpath none. src/main/ssh/config/ssh-known-hosts-source.test.ts: splits the space-separated list ssh -G prints on one line; expands ~ in a reported path; keeps a double-quoted path containing spaces whole; falls back to the default files when ssh -G reported nothing; drops an explicit none without falling back to the defaults; splits both known_hosts lists on one line each; reads StrictHostKeyChecking, including accept-new; reads HostKeyAlias and prefers it for the lookup; reads the non-default port; reports that the lookup came from an alias, so the port is not appended; reports no alias when the dialed host is used; unions the entries of every file; skips a path that is a directory; returns nothing when no file can be read; reports nothing unreadable when every file parses; does not count an empty file as unreadable; prefers HostKeyAlias over the resolved hostname; uses the dialed host, because a resolved hostname can just echo the label; still prefers an explicit HostKeyAlias over the dialed host; falls back to the dialed host when nothing was resolved; is rejoined when the split fragments resolve to nothing; leaves a single unresolvable path alone rather than inventing one; expands the __PROGRAMDATA__ token rather than passing it through as a path; does not rewrite a path that merely BEGINS with the token characters; keeps a drive-letter path with mixed separators intact; splits a spaced Windows home into the fragments the rejoin has to put back. src/main/ssh/config/ssh-known-hosts.test.ts: reads the algorithm from the blob rather than trusting the line; rejects %s; refuses a malformed blob: %s; uses the bare host on the default port; tries the bracketed form first and falls back to bare off-port; matches an exact line; matches case-insensitively; reports a changed key of the same type; reports unknown for %s; matches any host in a comma list; matches the bracketed form on its port; does not apply a bracketed entry to the default port; falls back to a bare line when off-port; never reports a change from the fallback pass; matches a bare literal on the default port; matches a bracketed literal on its port; reports a change for a literal; handles %s; still reports a change for a different key on a revoked host; skips a line carrying an unrecognised marker; does not report a change when only another key type is on file; matches within the same type; accepts CRLF terminators; skips blanks, comments and surrounding whitespace; accepts a hit regardless of which file holds it (%s); never brackets the alias, whatever the port; still brackets an ordinary host on a non-default port; matches a bare alias entry on a non-default port; still accepts the unmodified key field; accepts a hashed entry whose salt is a full SHA1 digest; still accepts the unmutated ssh-keygen -H vector; falls back to the bare host line on a non-default port, as ssh did; refuses a key type it has not seen for a host it knows, as ssh did. src/main/ssh/connection/ssh-connect-config.test.ts: returns undefined when no default keys exist; returns the first existing key file; probes regular and FIDO2 keys in stable default order; keeps a regular default ahead of a malformed FIDO2 default; uses target host/port/username; falls back to resolved config when target fields are empty; keeps imported endpoint fields as the fallback when ssh -G is unavailable; sets readyTimeout to CONNECT_TIMEOUT_MS; sets keepaliveInterval to 15s; enables keyboard-interactive auth so MFA challenges can be answered; uses agent auth when no explicit key and SSH_AUTH_SOCK is set; enables agent forwarding when OpenSSH config requests it and an agent is available; does not enable agent forwarding without a usable agent; resolves IdentityAgent SSH_AUTH_SOCK from the environment; uses the Windows OpenSSH agent pipe when no environment socket is available on Windows; wraps agent auth with IdentityFile filtering when IdentitiesOnly is enabled; applies a manual target IdentitiesOnly setting when ssh -G reports it off; includes unencrypted target.identityFile auth when an agent is available; defers encrypted target.identityFile auth when an agent is available; uses keyFile auth when target.identityFile is set and no agent is available; includes unencrypted resolved identityFile auth when an agent is available; defers a default-path resolved identityFile to the no-agent retry; offers the keyboard-interactive challenge when the agent defers no key; provides fallback key when no agent is available; can force private key inclusion for the post-agent fallback path; falls back to resolved proxyCommand; returns structured jump-host config for target.jumpHost; returns structured jump-host config for resolved proxyJump; returns undefined when no proxy is configured. src/main/ssh/connection/ssh-connection-auth-fallback.test.ts: resolves OpenSSH config using configHost when present; tries ssh-agent before reading an explicit private key; falls back to direct private key auth when agent auth fails; falls back to direct private key auth after too many agent authentication failures; answers bounded keyboard-interactive challenges such as Duo 2FA. src/main/ssh/connection/ssh-connection-store.test.ts: lists picker suppression aliases without consulting re-adoption tombstones; getTarget delegates to store; addTarget generates an id and persists; saves the CLI control switch as given, with no consent stamp; addTarget stamps source as manual by default; addTarget preserves an explicitly provided source; inserts a new config host stamped as ssh-config; asks the parser for all hosts — reconciliation happens in the store; updates an existing config-sourced target when the port changed; refreshes gssapiAuthentication on sync; never overwrites a manual target that owns the alias; does not rewrite an unchanged config-sourced target; returns empty array when nothing changed; tombstones a deleted config-sourced host so re-import skips it; suppresses a deleted manual target from config discovery; re-adding a deleted host reclaims its alias so sync stops suppressing it; editing a target reclaims its alias from the deleted set; records a tombstone when removing any user-facing target; re-adopts orphaned repos when the same host is re-added. src/main/ssh/connection/ssh-connection.test.ts: transitions to connected on successful connect; captures the negotiated SSH server key fingerprint; allows concurrent exec commands for ssh2 transport; removes startup listeners after ssh2 connect succeeds; scopes lifecycle events and pending handshake timers to one mock client; enables TCP_NODELAY on the new ssh2 client after a reconnect cycle; transitions through connecting → connected states; reports error state on connection failure; disconnect cleans up and sets state to disconnected; getTarget returns a copy of the target; getState returns a copy of the state; throws when connecting a disposed connection. src/main/ssh/orcad-migration-stuck-host-profiles.test.ts: never leaves a running terminal behind; is offered only when leaving saved state behind would clear the refusal. src/main/ssh/orcad-runtime-maintenance.test.ts: keeps a connect from retrying a build an explicit Update also failed; offers Force when Update was refused by policy; forwards anew after an activation, since the new slot may bind another port; passes a refusal through and leaves the server linked; forwards anew once a serving slot is restored, which may have bound another port. src/main/ssh/port-forwarding/ssh-port-forward.test.ts: reports the port the OS bound when an ssh2 forward asks for port 0; adds an ssh2 port forward and returns entry; adds a system SSH port forward when no ssh2 client is available; reports the port a system SSH forward resolved when asked for port 0; passes resolved OpenSSH config to system SSH port forwards; disposes a system SSH tunnel when removing the forward; lists forwards filtered by connectionId; removes a forward by id; awaits forward close when removing a forward for user actions; returns null when removing nonexistent forward; removes all forwards for a connection; dispose removes all forwards; stores label in the entry; preserves the existing id when updating a forward succeeds. src/main/ssh/ssh-channel-multiplexer.test.ts: sends a JSON-RPC request and resolves on response; assigns unique request IDs; sends notifications without expecting a response; dispatches incoming notifications to handler; sends one keepalive frame per cadence tick; drives keepalive sends AND dead-link detection from a single interval; resolves false immediately on a disposed mux; throws on request after dispose; reports isDisposed correctly; clears registered handlers on dispose; does not retain handlers registered after dispose. src/main/ssh/system-transport/system-ssh-forward-process.test.ts: spawns exactly the argv the orphan matcher is given, LogLevel=INFO included; runs port forwards unmultiplexed even when ssh_config disables muxing; preserves manual target port and identity options in the forwarded ssh command; does not spawn ssh when the requested local forward port is already in use; spawns ssh after the requested local forward port preflight succeeds; resolves startup after the grace period when ssh keeps running; sends SIGTERM then SIGKILL when the process does not exit; does not resolve stop until the process exits. src/main/ssh/system-transport/system-ssh-loopback-route.test.ts: dials the held forward ahead of the alias so OpenSSH keeps these over ssh_config; drops a manual target's own port and proxies, which would fight the route; never routes a forward, whose argv must stay stable for orphan recovery; %s → %s; opens the forward over the real proxy, then proves a login over the route; lands the forward on the sshd and known_hosts name the direct route resolved; holds the forward with server-alive probes; never lets the background forward or probe ask for credentials; starts no forward when admission refuses the route; stops the forward when sshd refuses logins over it; stops the whole forward process tree on Windows, once; reports a forward that already exited before anyone listened for it. src/main/startup/browser-process-user-agent-ordering.test.ts: explains admission refusal before a desktop launch exits; explains an orcad holding the profile instead of exiting silently; does not acquire profile admission for a duplicate launch; executes after the dev app name and before later preflight work; exits and releases admission after pending profile move recovery fails; keeps Linux background startup failures console-only and immediate; keeps Linux serve startup failures console-only and immediate. src/main/startup/quit/quit-path-durable-write-blocking.test.ts: store.flushAsync() issues no synchronous fs syscalls; store.flushAsync() also moves the active-view and github-cache sidecars off the thread; stats.flushAsync() issues no synchronous fs syscalls; Grok session cleanup issues no synchronous fs syscalls; stats.flushAsync() closes out live agents before it returns, not when its write lands; a stalled mount leaves the event loop live instead of parking it; store.flushAsync() is one idempotent final barrier; store.flushAsync() resolves instead of throwing when the profile mount rejects writes. src/main/stats/agent-session-transition-recorder.test.ts: counts a hook-only agent that never writes an agent-shaped terminal title; never opens a session from a replayed or disk-restored working status; still closes a live session when the terminating status arrives as a replay; counts one session per turn across repeated working/done cycles; closes an open session when its pane is torn down; bounds the mirror and closes what it evicts; stops the session when the main agent settles and only a watch loop holds the row working; keeps the session open while a subagent outlives the main agent, and closes it when the child settles; dates the stop by the evidence when a shell outlives the last subagent; opens a new session dated by the observation when the main agent resumes after monitoring; never dates a reopen before the close a fact-less OSC repaint caused; dates the first live row after a restored one by the evidence, not the restored clock; treats an unchanged answer as a snapshot, never a transition; opens only on a live working edge; closes only a session it opened. src/main/text-generation/commit-message-text-generation-model-discovery.test.ts: returns static catalog models without spawning for static agents; discovers dynamic models through the agent CLI; discovers dynamic models through the configured agent command override; falls back to static models when dynamic discovery returns no parseable models; falls back to the first discovered non-Pi model when its static default is unavailable. src/main/text-generation/pull-request-context.test.ts: does not fetch unrelated fork remotes before generating PR context; bounds exact remote-ref probe concurrency for a large remote list; reports no branch change because PR context preparation is read-only; keeps a remote-qualified base when the selected base includes the remote. src/main/text-generation/source-control-local-process.test.ts: asks the supervisor to stop and leaves its group to it; signals nothing once the supervisor has exited; stops an agent that ignores its stdin end through its supervisor. src/main/updater/linux-package/linux-package-update-recovery.test.ts: retains the downloaded package with the digest from the event metadata; requires a string downloaded path; accepts an uppercase extension; requires a non-empty string version; matches an absolute file URL by basename; accepts duplicate entries that agree on the digest; refuses an entry with a missing or non-string digest; refuses a digest that is not base64 at all; refuses a malformed percent-encoded URL; refuses a missing or non-array file list; tolerates a non-object event; replaces the retained artifact with a newer download; keeps the retained artifact when a new download has no usable digest; keeps the retained artifact when an unrelated download event arrives; clears the retained artifact; preserves the artifact across a recheck of the same version; clears the artifact when another version takes over; preserves the artifact for an unknown or empty version; is a no-op without a retained artifact; returns the built command and package file name for a verified package; reports missing without a retained artifact; propagates a package-manager discovery failure; rejects a well-formed digest that belongs to another file; never hashes a package that a malformed digest failed to arm; rejects a path outside the cache root; rejects the cache root itself; rejects a package in the updater directory itself; rejects a pending directory that is a symlink out of the cache; performs one hash pass for concurrent requests; revalidates once the in-flight pass settles; does not reuse an in-flight pass for a different artifact; starts a fresh proof when the same package is captured again; returns the verified package path; returns missing without a retained artifact. src/main/updater/releases/updater-prerelease-feed-readiness.test.ts: accepts a GitHub %s asset redirect as ready; reports a GitHub asset request error as unavailable; reports not-ready with a verified last-good tag when the newest assets are unavailable; does not classify a non-positive tag limit as feed unavailability; reproduces the v1.4.142 publishing incident as not-ready; reports transport failures as unavailable instead of not-ready; treats an explicit asset 404 as not-ready when another asset is unavailable; accepts absolute manifest asset URLs without rewriting them to release asset paths; returns not-ready with an older ready update as last-good while newest is publishing. src/main/updater/releases/updater-release-builds.test.ts: lists hourly builds from the dedicated repo, newest first; lists daily builds from the dedicated repo, newest first; separates stable from rc in the shared main repo; skips drafts and unparseable tags; keeps a composed release title and drops one that repeats the tag; hides builds that published no artifact for this platform; returns an empty list when no build has this platform artifact; keeps mac builds visible on macOS regardless of the Windows leg; resolves the platform installer download url; leaves the installer url null when the release published no installer; tolerates a release whose assets are missing or malformed; sends the local gh token as a bearer header when one is available; sends no authorization header when gh has no token; retries unauthenticated once when GitHub rejects the token; does not retry a 401 that was already unauthenticated; falls back to the per-IP bucket when the token is rate limited and tells the gh breaker; reports a secondary limit carrying retry-after without retrying or tripping the breaker; records a primary limit in the token scope %s; keeps %s authenticated when only %s is blocked; skips the token while its scope %s is blocked; surfaces a rate limit with its reset time and a sign-in hint when unauthenticated; omits the sign-in hint when the rate-limited request was authenticated; treats 429 as a rate limit; reports a 403 without rate-limit headers as a plain HTTP error; reports a missing hourly repo distinctly; is null when GitHub sent no reset; prefers retry-after over the primary reset epoch; falls back to the primary reset epoch when there is no retry-after; reads retry-after as seconds; reads retry-after as an HTTP date; pins an hourly tag at the hourly repo download path; pins a daily tag at the daily repo download path; pins a stable tag at the main repo download path; rejects a tag that is not a version. src/main/updater/updater.build-channel-selection.test.ts: uses the display label in the unsupported-platform %s pinned-build error; still pins %s from an unsigned Windows dev build; restores ordinary release checks after local build selection fails; opts into the RC channel when checkForUpdatesFromMenu is called with includePrerelease; pins the generic feed to a perf-tagged prerelease when requested; surfaces no-update feedback when no newer perf-tagged prerelease exists; keeps prerelease publishing-window misses on the generic retry path; leaves the feed URL alone for a normal user-initiated check. src/main/updater/updater.check-preflight.test.ts: ignores a stale update-not-available event after a new check starts preflight; deduplicates repeated manual checks while the immediate checking status is active. src/main/updater/updater.prerelease-fallback.test.ts: retries a prerelease check once against the previous feed tag when the manifest is missing; surfaces a promise-only prerelease fallback failure after the primary error event. src/main/updater/updater.publishing-window-feed.test.ts: repins the generic feed to the newest RC tag for a prerelease user; repins the generic feed to a newer stable tag for a prerelease user; pins a stable fallback when Atom has no newer prerelease; uses last-good concrete feed when a user-initiated check lands during publishing; can offer an older fully published update while the newest release is publishing; keeps background publishing-window fallback on the short retry cadence. src/main/updater/updater.quit-and-install.test.ts: surfaces an accepted retry before electron-updater emits download progress; defers quitAndInstall and runs its launcher within the launch PATH scope; runs pre-quit cleanup before local PTY cleanup during update install; keeps optional pre-quit cleanup fail-and-continue behavior by default; ignores duplicate quitAndInstall requests while the shared delay is pending; arms the forced-exit watchdog once the install commits; disarms the forced-exit watchdog when sync install error recovery keeps the app open. src/main/window/createMainWindow-markdown-editor-focus.test.ts: ignores traffic light sync IPC on non-macOS; still intercepts Cmd+Shift+B and Cmd+Alt+B when the markdown editor is focused; coerces non-boolean setMarkdownEditorFocused payloads to false; opens a table-aware context menu synchronously without a renderer query; builds the plain edit menu for the one right-click the renderer marks as plain text. src/main/window/createMainWindow-renderer-crash-recovery.test.ts: notifies the caller when the renderer process is gone; passes the renderer webContents id through crash recording and recovery callbacks; forwards expected renderer teardowns so the recorder can diagnose suppression; reloads the app shell after an unexpected renderer process loss; does not reload after renderer loss when recovery is disabled; stops auto-reloading after a rapid renderer crash loop trips the breaker. src/main/window/dashboard-popout-window.test.ts: updates the title after language initialization and unsubscribes when closed; lets its preload expose saved WSL hosts in serve-only, and passes nothing otherwise; creates a native-frame window with the shared preload and no webview surface; shows the window on ready-to-show; loads the prod file entry; loads the dev server URL when in dev; focuses the existing window instead of creating a second one; trusts only the live popout webContents; restores a minimized window when re-requested; restores valid persisted bounds; discards off-screen persisted bounds and falls back to defaults; persists bounds on resize after the debounce, guarding near-minimum sizes; keeps saving bounds after a vetoed quit and freezes them only on an allowed quit; closeDashboardPopout closes an open window; applies the persisted app zoom level on dom-ready; follows app zoom changes while open and unsubscribes on close; zoomDashboardPopoutIfFocused steps the popout zoom only while focused; handles the zoom-in chord via before-input-event and ignores other keys; handles mouse-wheel zoom requests outside before-input-event; respects zoom keybinding overrides for keyboard and mouse-wheel paths. src/main/window/dropped-path-preparation.test.ts: opens and stores each unique source once as drop size grows tenfold; passes ordinary paths without copying for %s; preserves ordinary successes and copied successes when another file fails; serializes copy requests from different owners; reports all lost paths without a shared reason when their failures differ. src/main/window/main-window-updater-deferred-setup.test.ts: releases all 64 completed fallback handles without repeating initialization; keeps setup after ready-to-show and releases its fallback afterward; preserves the 15-second fallback when first paint never arrives; preserves destroyed-window guard reads before and at the fallback deadline; keeps the exact %s error and releases its completed fallback; keeps the exact fallback %s error without retrying initialization; preserves pending manual setup after failed ready registration; keeps setup options, live store reads, nudge writes and required quit flush. src/main/worker-threads/worker-thread-request-queue.test.ts: posts one request at a time and in the order it was dispatched; fires a posted call at its own deadline; stops respawning and fails the rest of the queue once deaths hit the cap; clears the death count when a fresh burst starts from full idle. src/main/workspace-hooks/hooks-setup-runner-script.test.ts: builds Windows runners for newline-heavy scripts without line-array splitting; builds POSIX runners without regex-wide CRLF normalization; writes POSIX setup runners for shebang-declared scripts on native Windows paths; keeps batch setup scripts on cmd.exe when the terminal is Git Bash; hands a Git Bash pane a launch command MSYS cannot rewrite for a cmd runner; replays interpreter flags declared on the shebang line; preserves cmd.exe setup runner semantics for configured cmd users; keeps POSIX runner behavior on POSIX platforms; waits when either local or project policy requires completed setup; marks setup-runner terminals for the always-on credential guard; maps git-bash to POSIX setup launch metadata on Windows; falls back to the cmd runner when the git-bash setting has no installed Git Bash; keeps the cmd runner for a non-Git bash such as Cygwin; keeps the cmd runner for a bare bash whose flavor cannot be resolved; uses the POSIX runner for a bare bash that resolves to Git Bash; uses the POSIX runner for an extension-less Git Bash path; preserves the existing cmd runner for PowerShell terminals; preserves cmd setup compatibility when a Windows-host project has a WSL shell setting; classifies an installed explicit Git Bash executable as a POSIX runner; falls back to the cmd runner when the explicit Git Bash path no longer exists. src/main/workspaces/catalog/repo-git-remote-identity-enrichment.test.ts: schedules remote identity enrichment without blocking the caller; probes a drive project whose runtime is WSL %s; sweeps a local repo: %s; probes an SSH row that carries only executionHostId on its own host; keeps same-path rows on two different SSH hosts from sharing one backoff; coalesces concurrent probes for the same repo location; caches no-identity probes briefly so list calls do not retry every time; does not rewrite the no-remote marker on a later retry; resolves a settled no-remote repo once it gains a remote; does not re-probe a resolved identity before the refresh window elapses; re-probes a resolved identity after the refresh window elapses; bounds re-probes per sweep so a large repo list drains gradually; forgets a removed repo location so its deadline cannot outlive the repo; keeps a surviving repo backoff when a sibling repo is removed; probes again when the same location is re-added after its probe was retired; coalesces concurrent sweeps instead of running one pass per list call. src/main/workspaces/creation/preparation/worktree-create-preparation.test.ts: preserves create priority through claim probes on %s; starts the checkout only once the async workspace root resolves; still deduplicates when both callers await the same pending root lookup; namespaces native Windows preparation directories for long paths; deduplicates preparation for the same repo, base, runtime, and workspace root; does not spend a divergence walk when the base matches exactly; claims when the two sides spell the same ref differently; evicts a repo's own stale preparation before another repo's; still discards armed checkouts when the project runtime needs repair; routes preparation and finalization through the selected WSL runtime; retries stale cleanup after a transient listing failure; does not discard a detached worktree with caller-controlled preparation metadata; does not re-arm after an isolated create; re-arms a preparation once creates arrive in a burst; holds the re-arm checkout until the create runs the deferred thunk; starts one explicit prefetch after the create completes; does not re-arm when finalization failed; reports a pending create while a stale-cleanup scan is running. src/main/workspaces/removal/local-worktree-removal-recovery.test.ts: recovers Git for Windows partial filesystem deletion failures; finishes a clean non-force deletion after Git leaves a Windows directory behind; recovers localized Windows failures after Git already removed the registration; does not recover unrelated localized failures while Git still owns the row; finishes a macOS deletion Git deregistered before failing with Directory not empty; leaves an orphaned checkout to the orphan cleanup instead of deleting it; prunes and strictly verifies on the selected WSL host without deleting files or branches. src/main/workspaces/removal/worktree-removal-safety.test.ts: rejects deleting a worktree that contains another registered worktree; rejects POSIX home directories even when host homedir has a different path shape; refuses a home the host reported even when no path rule recognises it; refuses the Windows host's home directory from a POSIX client; refuses the Windows host's home directory even when the host reported nothing. src/main/wsl/discovery/distro-catalog.test.ts: retries asynchronous discovery after a transient wsl.exe failure; retries synchronous discovery after a transient wsl.exe failure; retries asynchronous discovery after wsl.exe reports no distros yet; retries synchronous discovery after wsl.exe reports no distros yet; bounds the wsl.exe spawn rate while no distros are installed; bounds the asynchronous spawn rate while no distros are installed; still caches a non-empty distro list for the process lifetime; re-probes a docker-desktop-only machine once a user distro can appear; backs off while the list keeps coming back empty; backs off after repeated distro-list failures; caps the empty-list backoff at five minutes; keeps reporting an empty result after it goes stale; keeps reporting a known distro list indefinitely; bounds cached homes while retaining the most recently used distros; still lists distros, and an excluded distro home probe fails like any other probe; gives the async home probe time for a cold boot instead of killing it at 5 s; retries availability after a probe timeout instead of latching false; bounds the probe rate while the failure window is open; caches a successful probe for the process lifetime; names an explicit spawn directory, so no deleted cwd can read as ENOENT; shares one wsl.exe spawn between concurrent probes; keeps reporting the last observed answer after it goes stale; holds a definitive failure far longer than a timeout when %s; backs off after repeated probe failures; re-probes a seeded retryable failure once its window lapses; re-probes availability once a distro list succeeds after %s; does not drop an availability failure for an empty distro list; keeps reporting unavailable off Windows without probing. src/main/wsl/discovery/wsl-distro-holder.test.ts: runs a busybox-safe loop through --exec and names the distro; spawns a detached-style holder, records it, and waits for the distro to run; a holder exiting with code 1 notes its distro may have stopped (E9); rejects a pid whose command line is not our holder for that distro; cleans up stale and corrupt records; keeps a holder and its record when release cannot verify it; reports unverifiable when the command line is unreadable even with a matching pid; release kills a verified holder and removes its record; release never kills a pid that is no longer our holder; reports a spawned holder exit so liveness can re-check. src/main/wsl/filesystem/wsl-guest-drive-mountinfo.test.ts: skips 9p shares that are not drvfs, and malformed lines. src/main/wsl/orcad/wsl-orcad-deploy-app-update.test.ts: plans the update on the census the running orcad reports, not an unknown one; forwards force to the update planner; does not serve the incumbent past a refusal that is not a wait; does not re-activate the build this host rolled back from; labels the launch with the build the adoption found serving, not the stale record; defers while another run holds the host fence; lets an explicit force deploy past both skips; names a newer host refusal apart from a rolled-back one, so only Force is offered; updates an older host as before, without asking to adopt it; never asks the incumbent when it already runs this build. src/renderer/src/components/agent/agent-launch-picker-acquisition.test.ts: keeps a completed listing usable when the host rejects the wait parameter. src/renderer/src/components/feature-tips/feature-tip-telemetry.test.ts: delivers to chat-default users in all targeting while holding out not-switched users. src/renderer/src/components/native-chat/composer/pickers/use-native-chat-side-chat-command.test.ts: completes the row on a pick, even when the command is already typed; ignores a pick of the row while side chats are unavailable. src/renderer/src/components/native-chat/media/images/native-chat-image-copy.test.ts: explains a too-large refusal that arrives through the IPC error prefix. src/renderer/src/components/native-chat/messages/actions/native-chat-fork-launch.test.ts: keeps the question with the caller and says once that it never went; tells the user to check the side chat when delivery is unconfirmed; reports a delivered question as sent, silently. src/renderer/src/components/native-chat/messages/actions/native-chat-side-chat-fork-point.test.ts: decides each of %i tool-only turns once per layout, and nothing without a listener. src/renderer/src/components/native-chat/messages/actions/use-native-chat-fork.test.tsx: places a side chat whose question was %s: %o. src/renderer/src/components/terminal-pane/provider-notices/agent-presence/agent-terminal-presence.test.ts: calls back once for %s; stops watching once unsubscribed. src/renderer/src/components/terminal-pane/pty-connection/emulation/pty-connection-mode2031-raw-subscription.spec.ts: records a mode 2031 subscribe from the raw chunk boundary without answering. src/renderer/src/components/terminal-pane/pty-connection/input/pty-connection-deferred-input-forwarding.spec.ts: routes terminal input through deferPtyInput so the host can withhold it; forwards terminal input directly when the host supplies no deferPtyInput. src/renderer/src/components/terminal-pane/pty-connection/notifications/pty-connection-terminal-bell-attention.spec.ts: forwards onBell to pane attention and notification policy; hands the bell to the attention module with its pane even when pane attention is off. src/renderer/src/components/terminal-pane/pty-connection/output-restore/pty-connection-dropped-output-snapshot-healing.spec.ts: repaints from the main-owned snapshot when main drops pending output at the cap. src/renderer/src/components/terminal-pane/pty-connection/output/pty-connection-fresh-native-follow-reset.spec.ts: retries a fresh-spawn native follow reset when renderer dimensions return. src/renderer/src/components/terminal-pane/pty-connection/spawn/unbound-pane-spawn-recovery.test.ts: requests a tab remount for an unbound spawn; leaves recovery to a remote transport that armed its own retry; requests a remount when the transport reports no recovery of its own; settles the spawn as failed before requesting a remount; prefers the restored leaf id when reporting the anomaly; falls back to the pane leaf id when no restored leaf exists. src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-foreground-agent.test.ts: remove only literal registration import/call pairs import { registerPtyConnectionDeferredInputForwardingTests } from '../input/pty-connection-deferred-input-forwarding.spec', registerPtyConnectionDeferredInputForwardingTests(); no additional authored credit. src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-hidden-replay.test.ts: structural support changed only; no new authored removal credit. src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-input-and-repaint.test.ts: remove only literal registration import/call pairs import { registerPtyConnectionFreshNativeFollowResetTests } from '../output/pty-connection-fresh-native-follow-reset.spec', registerPtyConnectionFreshNativeFollowResetTests(); no additional authored credit. src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-spawn-and-startup.test.ts: remove only literal registration import/call pairs import { registerPtyConnectionTerminalBellAttentionTests } from '../notifications/pty-connection-terminal-bell-attention.spec', registerPtyConnectionTerminalBellAttentionTests(); no additional authored credit. src/renderer/src/components/terminal-pane/pty-transport/delivery/pty-dispatcher-delivery-interest.test.ts: registers interest on the first sidecar and releases on the last unsubscribe; releases sidecar interest only once for repeated unsubscribes; does not let an eager pre-mount buffer defeat hidden delivery gating; keeps eager buffers outside delivery interest when the PTY exits before mount; lets a sidecar exclusively own interest while an eager buffer overlaps. src/renderer/src/components/terminal-pane/pty-transport/delivery/pty-dispatcher-push-reattach.test.ts: blackholed delivery reproduces the wedge: no handler dispatch, no ACK ever. src/renderer/src/components/terminal-pane/pty-transport/ipc/ipc-pty-opencode-model.test.ts: sends preferences to the execution host and attaches only to its returned PTY; preserves one operation ID across a lost reply and reconnect; preserves ordinary raw startup and captured-session attachment. src/renderer/src/components/terminal-pane/pty-transport/output/pty-side-effect-pending-census.test.ts: sums registered gauges and drops disposed ones; ignores a repeated dispose instead of dropping another gauge; stops tracking new gauges past the cap so a missed dispose cannot grow it; tracks a live output processor queue through enqueue, drain, and dispose; excludes cleared consumed entries from retention after a bounded drain. src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-prebind-discard-notice.test.ts: says nothing once the pane is closed. src/renderer/src/components/terminal-pane/render-health/terminal-webgl-atlas-recovery.test.ts: refreshes after each scheduled atlas reset; falls back to a timeout when animation frames are unavailable; continues recovery when a manager throws after scheduling. src/renderer/src/components/terminal-pane/snapshots/pty-buffer-serializer.test.ts: tracks a remounted quiet title source even when the stale owner unregisters later; preserves the renderer-parsed output sequence in the response; preserves a pending escape tail for snapshot replay; routes a host's Reset Terminal request to the pane that owns the PTY. src/renderer/src/components/terminal-pane/snapshots/saved-terminal-output-replay.test.ts: marks panes with restored scrollback for fresh-shell viewport blanking. src/renderer/src/components/terminal-pane/viewport/split-right-white-screen.test.ts: recovers a 0x0 spawn whose container is briefly unmeasurable then settles. src/renderer/src/components/terminal-pane/viewport/visibility/stale-document-visibility.test.ts: removes global listeners when the last pane unregisters. src/renderer/src/components/ui/security-surface-theme-contrast.test.tsx: keeps trust text readable on a host-owned dialog surface; applies the hostile surface to ordinary dialogs, so the case above is exercised. src/renderer/src/components/use-terminal-parking-pass.test.tsx: does not queue a render when a pass produces the same parking sets; keeps parking transitions working under StrictMode. src/renderer/src/hooks/client-workspace-view-report.test.ts: reports an away user’s screen with no presence, and renews as soon as input returns; retries a failed send, empty or not, no sooner than a refresh. src/renderer/src/hooks/ipc-events/terminal-request-ipc-bridge-placement.test.ts: is new work in a workspace with no live terminal: its first bind moves it up in Recent. src/renderer/src/hooks/use-bundled-skill-install.test.tsx: ignores a repeated click while the install runs; retries the last requested skills on this host. src/renderer/src/hooks/useGitHubSlugMetadata.assignee-reuse.test.tsx: adds seeds to a successful empty list but not to loading or failed results; retains seed projection after expiry and fetches fresh metadata for a new reader. src/renderer/src/hooks/useSettingsNavigationMetadata.language-switch.test.tsx: retranslates the settings nav labels when the UI language changes live. src/renderer/src/lib/codex-stale-pane-account-identity.test.ts: keeps the prompt when the roster read fails and every label collapses. src/renderer/src/lib/image-blob-png.test.ts: downscales a pasted image above the pixel cap; releases its temporary URL on a decode error; releases its temporary URL when PNG encoding fails or exceeds the byte cap. src/renderer/src/lib/react-commit-cascade-store-write-samples.test.ts: names callers above the frame the write came from; restores Error.stackTraceLimit when the capture throws; stops collecting after the sample cap when captureStackTrace is unavailable; caps reported changed keys from a single wide write. src/renderer/src/lib/work-items/linear/linear-issue-context-snapshot.test.ts: applies the rendered text hard cap last and reserves the final overflow marker. src/renderer/src/runtime/rpc/targets/use-worktree-runtime-target.test.ts: keeps the runtime target identity stable across unrelated store writes. src/renderer/src/runtime/source-control/git/runtime-git-branch-client.test.ts: keeps creation available on old hosts and explains the missing operation; preserves landed warnings and direct collision recovery data; propagates remote discovery failure instead of a false empty page. src/renderer/src/runtime/sync-runtime-graph/sync-runtime-graph-key-reuse.test.ts: reuses unchanged worktree tab projections when one worktree title changes; bumps only the changed worktree version when a sibling stays byte-identical. src/renderer/src/store/repos/repos-catalog-request-coordination.test.ts: shares safe-auto fork sync cooldown attempts across store instances. src/renderer/src/store/slices/tabs/hydrated-workspace-reconciliation-batch.test.ts: collapses a 193-workspace hydration to one store write; leaves the store byte-identical to the per-workspace path; re-reconciling the batched result is a no-op, as it is for the per-workspace path. src/renderer/src/store/slices/tabs/hydrated-workspace-reconciliation-fixture.ts: dedicated zero-case helper retirement; no authored test credit. src/renderer/src/store/slices/worktrees/metadata/update-worktrees-meta.test.ts: persists each batch metadata update to its workspace owner. src/renderer/src/store/slices/worktrees/metadata/worktree-meta-rename.test.ts: dismisses a rename error on its capable remote owner; clears pending first-agent rename when the title is updated. src/renderer/src/store/terminals/store-sleep-shutdown-order.test.ts: orders capture, guards, handlers, main's stop, commit, cleanup, and exit settlement. src/renderer/src/store/terminals/terminal-first-tab-publication.test.ts: publishes the first terminal and root tab group atomically; publishes the first terminal and root tab group atomically, then marks it starting in main. src/renderer/src/store/terminals/workspace-terminal-reconnect-hydration.test.ts: restores daemon terminal wake hints before declaring the session ready. src/renderer/src/web/web-runtime-connection-auth-framing.test.ts: keeps legacy framing for sockets that cannot absorb a refused attempt. src/shared/terminal/appearance/terminal-font-weight-windows-regular.test.ts: moves the old Windows default to Regular once and queues the notice; keeps custom weights and fresh profiles without a notice; leaves macOS and Linux profiles untouched. tests/e2e/claude-agent-row-view.unit.test.ts: updates a mounted compact workflow description through the subscribed production parent; bounds upstream active-SSH visits for $population records, fallback=$fallback; derives %s destinations without requiring type or a workflow child ID; terminates a $path parent $status with no third request or intermediate toast; does not recover a first child %s. tests/e2e/cross-version-wire/agent-permission-legacy-spawn.unit.test.ts: keeps exact legacy ordering when structured creation cannot express afterTabId; can create a terminal without selecting the target worktree; preserves the legacy fresh-agent path when host authority is unavailable. tests/e2e/paired-instant-browser-tab.spec.ts: shows a paired browser tab before its create RPC resolves, then keeps it as one tab; adopts a paired browser tab without rebuilding its chrome. tests/e2e/session-tabs-decorative-title-fanout.unit.test.ts: bounds host publication and renderer application across remote worktrees; globally caps decorative freshness while every working status stays fresh. tests/e2e/session-tabs/mobile-session-active-performance.unit.test.ts: bounds saved-order publication selection at N and 10N without rematerializing navigation; bounds admitted placement reconciliation during supported reads at N and 10N; bounds selection input visits at 50 and 500 rows, including unavailable-target projection; keeps publication scoped to a changed worktree and shared by listeners at N and 10N; bounds preserved-chat merge and runtime-owned hydration at N and 10N (legacy=%s); bounds actual native retirement and shared delivery work at N and 10N. tests/e2e/session-tabs/mobile-session-active-publication.unit.test.ts: does not substitute a preserved row for an unavailable explicit legacy selection; resolves main-layout group selection %s (omitted group=%s). tests/e2e/structured-agent-session-launch-admission.unit.test.ts: opens a direct caller's terminal with the caller's own CLI arguments on a decline; opens the chat on the server that admitted it; opens only the terminal when this machine declines, without a notice; opens a new agent tab's terminal with the typed prompt when the caller names none; opens the chat, not a terminal, when this machine cannot resolve a new workspace yet; keeps its terminal fallback, as before. tests/e2e/structured-attention-prompt-relay.unit.test.ts: settles once per relayed prompt, not on every later status update. tests/e2e/structured-attention-read-retirement.unit.test.ts: Mark read on a chat never opened withdraws the prompt alert it surfaced; Mark read on a chat hidden since an earlier view covers the newer prompt it surfaced; reads a newly accepted visible prompt without sending an RPC for text-only updates. tests/e2e/structured-chat-tab-lifetime-cost.unit.test.ts: counts real raw stores, unrelated window publications and captured cleanup at %i chats; counts complete supplied-member sweep and separate final retirement at S=%i N=%i. tests/e2e/structured-session-workspace-authority.unit.test.ts: bounds actual target reveal across %i repo roots and metadata aliases. tests/e2e/terminal-agent-pane-launch.unit.test.ts: starts the host's planned command once, at the pane's size, and says it started; plans a pane whose prompt the window submits without the startup update prompt; preserves caller shell-ready delivery only for a connection-scoped adapter ($connectionId); starts no shell on plan failure (original failure: $returnLaunchFailure); cannot confirm a start whose provider answer was lost. tests/e2e/terminal-parked-memory.spec.ts: releases renderer terminal memory when hidden tabs park. tests/e2e/terminal-shortcuts.spec.ts: @headful Codex-like background output stays visible without disabling WebGL in auto mode; Shift with Russian layout text reaches the PTY as Cyrillic under kitty keyboard reporting. tests/e2e/terminal-split-activation-latency.spec.ts: records attributed CWD, spawn, bind, fixture-ready, input, and echo phases. tests/e2e/terminal-split-activation-latency-phases.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-split-activation-latency-main-probe.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-split-activation-latency-artifact.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-split-activation-latency-report.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-tab-switch-visual-restore.spec.ts: restores skipped hidden Grok output on light tab resume. tests/e2e/web-runtime-session-terminal-host-authority.unit.test.ts: delivers independent %s effort through the real prepared web callback; keeps same-ID local and runtime worktrees on the selected runtime owner; creates paired web agents through host authority so activation is mirrored; delivers generated continuation context after host-authoritative creation; seeds the chat composer for a draft that rode in on the launch command. Current changed source bytes, wrapper registrations and reduced selectors have no fresh execution or red/green credit. Unrelated surviving receipt objects keep their original narrow scope; current literal graph reachability does not establish historical import closure or equivalent replacement coverage. Current conservative command/import graph reaches src/main/persistence/worktree-sort-order-write.test.ts, src/main/runtime/construction/orca-runtime.test.ts, src/main/runtime/orca-runtime-tests/targeting-and-resilience.spec.ts, src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-04.spec.ts, src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-03.spec.ts, src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-02.spec.ts, src/main/runtime/orca-runtime-tests/gitlab-and-pr-bases-part-02.spec.ts, src/main/runtime/orca-runtime-tests/gitlab-and-pr-bases.spec.ts, src/main/runtime/orca-runtime-tests/worktree-setup-and-startup-part-03.spec.ts, src/main/runtime/orca-runtime-tests/worktree-setup-and-startup.spec.ts, src/main/runtime/orca-runtime-tests/terminal-sleep-and-teardown-part-02.spec.ts, src/main/runtime/orca-runtime-tests/mobile-summaries-part-03.spec.ts, src/main/runtime/orca-runtime-tests/mobile-summaries-part-02.spec.ts, src/main/runtime/orca-runtime-tests/mobile-summaries.spec.ts, src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-03.spec.ts, src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-02.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-13.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-12.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-11.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-10.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-08.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-07.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-05.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-04.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-02.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-05.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-04.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-03.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-02.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status.spec.ts, src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-07.spec.ts, src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-04.spec.ts, src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-03.spec.ts, src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-02.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-12.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-11.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-10.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-09.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-08.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-07.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-06.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-03.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness.spec.ts, src/main/runtime/orca-runtime-tests/headless-snapshots.spec.ts, src/main/runtime/orca-runtime-tests/terminal-side-effect-facts-part-02.spec.ts, src/main/runtime/orca-runtime-tests/local-worktree-creation-part-02.spec.ts, src/main/runtime/orca-runtime-tests/local-worktree-creation.spec.ts, src/main/runtime/integrations/skill/orca-runtime-skill-recovery.test.ts, src/main/runtime/orchestration/mailbox/routing/orchestration-mailbox-detached-routing.spec.ts, src/main/runtime/orchestration/mailbox/terminal-pointer/orchestration-mailbox-delivery.test.ts, src/main/runtime/orchestration/mailbox/terminal-pointer/orchestration-mailbox-pointer-cli-command.spec.ts, src/main/runtime/rpc/methods/terminal/terminal-multiplex-link-estimate.test.ts, src/main/runtime/rpc/methods/terminal/terminal-snapshot-scrollback-rows.test.ts, src/main/runtime/rpc/server/pairing/network/windows-default-route-interfaces.test.ts, src/main/runtime/rpc/server/pairing/pairing-offer-owner-cli.test.ts, src/main/runtime/rpc/terminal-multiplex-end-verdict.test.ts, src/main/runtime/session-graph/headless-runtime-graph.spec.ts, src/main/runtime/session-graph/runtime-graph-standing-contracts.test.ts, src/main/runtime/session-tabs/structured-conversation-tab-title.test.ts, src/main/runtime/session-tabs/terminal-close-observed-exit.test.ts, src/main/runtime/terminal/identity/terminal-identity-probe.test.ts, src/main/runtime/terminal/layout/paired-desktop-server-writes-layout-owner.test.ts, src/main/runtime/terminal/lifecycle/pty-lifecycle-generation-retention.test.ts, src/main/runtime/terminal/lifecycle/run-terminal-user-open.test.ts, src/main/runtime/terminal/surface/terminal-surface-replies-window.test.ts, src/main/runtime/terminal/surface/workspace-layout-projection.test.ts, src/main/runtime/terminal/view/terminal-subscribe-exit-waiter-leak.test.ts, src/main/runtime/terminal/wait/orca-runtime-exact-worker-provider-session.test.ts, src/main/runtime/terminal/wait/prime-agent-screen-readiness-transcripts.test.ts, src/main/runtime/workspaces/startup/runtime-worktree-agent-startup.test.ts, src/main/runtime/workspaces/startup/runtime-worktree-startup-readiness.test.ts, src/main/skills/skill-install-transaction.test.ts, src/main/speech/model-manager-download-resume.test.ts, src/main/ssh/config/ssh-config-parser.test.ts, src/main/ssh/config/ssh-known-hosts-source.test.ts, src/main/ssh/config/ssh-known-hosts.test.ts, src/main/ssh/connection/ssh-connect-config.test.ts, src/main/ssh/connection/ssh-connection-auth-fallback.test.ts, src/main/ssh/connection/ssh-connection-store.test.ts, src/main/ssh/connection/ssh-connection.test.ts, src/main/ssh/orcad-migration-stuck-host-profiles.test.ts, src/main/ssh/orcad-runtime-maintenance.test.ts, src/main/ssh/port-forwarding/ssh-port-forward.test.ts, src/main/ssh/ssh-channel-multiplexer.test.ts, src/main/ssh/system-transport/system-ssh-forward-process.test.ts, src/main/ssh/system-transport/system-ssh-loopback-route.test.ts, src/main/startup/browser-process-user-agent-ordering.test.ts, src/main/startup/quit/quit-path-durable-write-blocking.test.ts, src/main/stats/agent-session-transition-recorder.test.ts, src/main/text-generation/commit-message-text-generation-model-discovery.test.ts, src/main/text-generation/pull-request-context.test.ts, src/main/text-generation/source-control-local-process.test.ts, src/main/updater/linux-package/linux-package-update-recovery.test.ts, src/main/updater/releases/updater-prerelease-feed-readiness.test.ts, src/main/updater/releases/updater-release-builds.test.ts, src/main/updater/updater.build-channel-selection.test.ts, src/main/updater/updater.check-preflight.test.ts, src/main/updater/updater.prerelease-fallback.test.ts, src/main/updater/updater.publishing-window-feed.test.ts, src/main/updater/updater.quit-and-install.test.ts, src/main/window/createMainWindow-markdown-editor-focus.test.ts, src/main/window/createMainWindow-renderer-crash-recovery.test.ts, src/main/window/dashboard-popout-window.test.ts, src/main/window/dropped-path-preparation.test.ts, src/main/window/main-window-updater-deferred-setup.test.ts, src/main/worker-threads/worker-thread-request-queue.test.ts, src/main/workspace-hooks/hooks-setup-runner-script.test.ts, src/main/workspaces/catalog/repo-git-remote-identity-enrichment.test.ts, src/main/workspaces/creation/preparation/worktree-create-preparation.test.ts, src/main/workspaces/removal/local-worktree-removal-recovery.test.ts, src/main/workspaces/removal/worktree-removal-safety.test.ts, src/main/wsl/discovery/distro-catalog.test.ts, src/main/wsl/discovery/wsl-distro-holder.test.ts, src/main/wsl/filesystem/wsl-guest-drive-mountinfo.test.ts, src/main/wsl/orcad/wsl-orcad-deploy-app-update.test.ts, src/renderer/src/components/agent/agent-launch-picker-acquisition.test.ts, src/renderer/src/components/feature-tips/feature-tip-telemetry.test.ts, src/renderer/src/components/native-chat/composer/pickers/use-native-chat-side-chat-command.test.ts, src/renderer/src/components/native-chat/media/images/native-chat-image-copy.test.ts, src/renderer/src/components/native-chat/messages/actions/native-chat-fork-launch.test.ts, src/renderer/src/components/native-chat/messages/actions/native-chat-side-chat-fork-point.test.ts, src/renderer/src/components/native-chat/messages/actions/use-native-chat-fork.test.tsx, src/renderer/src/components/terminal-pane/provider-notices/agent-presence/agent-terminal-presence.test.ts, src/renderer/src/components/terminal-pane/pty-connection/emulation/pty-connection-mode2031-raw-subscription.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/input/pty-connection-deferred-input-forwarding.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/notifications/pty-connection-terminal-bell-attention.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/output-restore/pty-connection-dropped-output-snapshot-healing.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/output/pty-connection-fresh-native-follow-reset.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/spawn/unbound-pane-spawn-recovery.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-foreground-agent.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-hidden-replay.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-input-and-repaint.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-spawn-and-startup.test.ts, src/renderer/src/components/terminal-pane/pty-transport/delivery/pty-dispatcher-delivery-interest.test.ts, src/renderer/src/components/terminal-pane/pty-transport/delivery/pty-dispatcher-push-reattach.test.ts, src/renderer/src/components/terminal-pane/pty-transport/ipc/ipc-pty-opencode-model.test.ts, src/renderer/src/components/terminal-pane/pty-transport/output/pty-side-effect-pending-census.test.ts, src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-prebind-discard-notice.test.ts, src/renderer/src/components/terminal-pane/render-health/terminal-webgl-atlas-recovery.test.ts, src/renderer/src/components/terminal-pane/snapshots/pty-buffer-serializer.test.ts, src/renderer/src/components/terminal-pane/snapshots/saved-terminal-output-replay.test.ts, src/renderer/src/components/terminal-pane/viewport/split-right-white-screen.test.ts, src/renderer/src/components/terminal-pane/viewport/visibility/stale-document-visibility.test.ts, src/renderer/src/components/ui/security-surface-theme-contrast.test.tsx, src/renderer/src/components/use-terminal-parking-pass.test.tsx, src/renderer/src/hooks/client-workspace-view-report.test.ts, src/renderer/src/hooks/ipc-events/terminal-request-ipc-bridge-placement.test.ts, src/renderer/src/hooks/use-bundled-skill-install.test.tsx, src/renderer/src/hooks/useGitHubSlugMetadata.assignee-reuse.test.tsx, src/renderer/src/hooks/useSettingsNavigationMetadata.language-switch.test.tsx, src/renderer/src/lib/codex-stale-pane-account-identity.test.ts, src/renderer/src/lib/image-blob-png.test.ts, src/renderer/src/lib/react-commit-cascade-store-write-samples.test.ts, src/renderer/src/lib/work-items/linear/linear-issue-context-snapshot.test.ts, src/renderer/src/runtime/rpc/targets/use-worktree-runtime-target.test.ts, src/renderer/src/runtime/source-control/git/runtime-git-branch-client.test.ts, src/renderer/src/runtime/sync-runtime-graph/sync-runtime-graph-key-reuse.test.ts, src/renderer/src/store/repos/repos-catalog-request-coordination.test.ts, src/renderer/src/store/slices/tabs/hydrated-workspace-reconciliation-batch.test.ts, src/renderer/src/store/slices/tabs/hydrated-workspace-reconciliation-fixture.ts, src/renderer/src/store/slices/worktrees/metadata/update-worktrees-meta.test.ts, src/renderer/src/store/slices/worktrees/metadata/worktree-meta-rename.test.ts, src/renderer/src/store/terminals/store-sleep-shutdown-order.test.ts, src/renderer/src/store/terminals/terminal-first-tab-publication.test.ts, src/renderer/src/store/terminals/workspace-terminal-reconnect-hydration.test.ts, src/renderer/src/web/web-runtime-connection-auth-framing.test.ts, src/shared/terminal/appearance/terminal-font-weight-windows-regular.test.ts, tests/e2e/claude-agent-row-view.unit.test.ts, tests/e2e/cross-version-wire/agent-permission-legacy-spawn.unit.test.ts, tests/e2e/paired-instant-browser-tab.spec.ts, tests/e2e/session-tabs-decorative-title-fanout.unit.test.ts, tests/e2e/session-tabs/mobile-session-active-performance.unit.test.ts, tests/e2e/session-tabs/mobile-session-active-publication.unit.test.ts, tests/e2e/structured-agent-session-launch-admission.unit.test.ts, tests/e2e/structured-attention-prompt-relay.unit.test.ts, tests/e2e/structured-attention-read-retirement.unit.test.ts, tests/e2e/structured-chat-tab-lifetime-cost.unit.test.ts, tests/e2e/structured-session-workspace-authority.unit.test.ts, tests/e2e/terminal-agent-pane-launch.unit.test.ts, tests/e2e/terminal-parked-memory.spec.ts, tests/e2e/terminal-shortcuts.spec.ts, tests/e2e/terminal-split-activation-latency.spec.ts, tests/e2e/terminal-split-activation-latency-phases.ts, tests/e2e/terminal-split-activation-latency-main-probe.ts, tests/e2e/terminal-split-activation-latency-artifact.ts, tests/e2e/terminal-split-activation-latency-report.ts, tests/e2e/terminal-tab-switch-visual-restore.spec.ts, tests/e2e/web-runtime-session-terminal-host-authority.unit.test.ts. Unresolved local/computed edges remain explicit debt.", "Audit2095 current-parent preparation2126 (approved2092/2093 removals; source refresh2109/independent2112): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-sleep-and-teardown-part-03.spec.ts: rejects exact terminal stop for multiple expected PTYs before stopping anything; omits stale renderer leaves when fresh PTY liveness is required; omits unbound renderer placeholders when fresh PTY liveness is required; captures the event method before the nested sleep receipt changes it; returns the close receipt without requiring a PTY stop; uses the same stop result as an empty explicit stop. src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-04.spec.ts: builds a compact worktree summary from persisted and live runtime state; reads the linked-PR state from the renderer repoId-keyed GitHub cache; carries persisted worktree host ownership in mobile summaries; emits only instance- and boundary-validated lineage parents in mobile summaries; resolves WSL platforms only for repos represented in mobile summaries; omits worktrees hidden by the host visibility policy from mobile summaries; applies linked-checkout source visibility to mobile summaries. src/main/runtime/orca-runtime-tests/hooks-and-hosted-review-part-02.spec.ts: routes runtime GitHub PR details and actions through the selected WSL project runtime; passes SSH connection context through hosted review creation flows; routes local WSL project hosted review flows through runtime git options; treats SSH worktree drift as unknown without local git probes. src/main/runtime/orca-runtime-tests/terminal-listing.spec.ts: emits one mobile session terminal tab per live PTY even if two tabs resolve to it; resolves projected split authority from the parent layout PTY binding; keeps explicit-id terminal lists from resolving all worktrees; keeps explicit-id cold-cache terminal lists from classifying unrelated same-repo worktrees; keeps unknown bare terminal-list ids on the no-scan exact-id path; matches explicit-id cwd PTYs for folder workspace instance IDs; routes PTY output through the PTY leaf index in large terminal graphs. src/main/runtime/orca-runtime-tests/window-authority.spec.ts: claims the first window as authoritative and ignores later windows; transfers authority from the headless sentinel to the first real window; relays terminal browser launches only while headless owns the graph; marks live headless PTYs for renderer reattach before desktop promotion; reports the activation gate state while no desktop window is available; bumps the epoch and enters reloading when the authoritative window reloads; can mark the graph ready for the authoritative window; drops back to unavailable and clears authority when the window disappears; hands the graph back to headless when a promoted serve window closes (#27292); keeps a crashed promoted serve renderer on its own window while the window is open; stays unavailable during initial loads before a graph is published. src/main/runtime/orchestration/mailbox/routing/orchestration-mailbox-routing-races.spec.ts: stops paged routing when the check is cancelled; keeps an unfiltered redrive unrestricted when reservations merge; keeps an unfiltered parked delivery unrestricted when reservations merge; uses a bounded indexed pane lookup for reminted Dispatch identity. src/main/runtime/workspaces/removal/worktree-teardown-unstopped-pty.test.ts: removes the reported wedged automation workspace without --force; clears PTY state once a failed stop is proven to have exited; names only the PTYs that are actually live, not every failed stop; warns with the specific sweep failure, not the generic deadline; still fails closed on a sweep-level failure without force. src/main/skills/bundled-skill-install/layouts.test.ts: links into a project only where its agent folder already exists; reads a pending cloud transaction as busy and a cloud receipt as another skill. src/main/skills/skill-upload-session-service.test.ts: joins concurrent disposal callers through exact cleanup completion. src/main/source-control/hosted-review-creation.test.ts: creates the pull request after fresh main-process validation passes; routes local WSL git and GitHub review creation through the selected runtime; creates a pull request on a GitHub Enterprise Server remote (#8312); creates a GitLab merge request after fresh main-process validation passes; creates an Azure DevOps pull request after fresh main-process validation passes; creates a Gitea pull request after fresh main-process validation passes. src/main/source-control/hosted-review-execution-host-routing.test.ts: reads a GitHub review on the SSH host a row names only in executionHostId; reads a GitLab review on the executionHostId host when connectionId names a different one; does not dial this client for a runtime row whose nested target shares a local name; keeps a runtime row with no nested target on this process (self-addressed stamp); keeps a legacy connectionId-only SSH row on its target; does not share a cached answer between two rows at one path on different hosts; creates a Bitbucket review locally for a runtime row rather than dialing its nested target; reports creation eligibility from the executionHostId-only SSH host; resolves the GitHub slug for a repo listing on the row it is asked about; refuses to dial a runtime host from this process; answers with the SSH target for both spellings and local otherwise; reads a runtime stamp on a row in this store as self-addressing, not a second machine. src/main/source-control/hosted-review-inflight-invalidation.test.ts: keeps other hosts, paths and their pending promises isolated. src/main/speech/model-manager.test.ts: requires pinned, internally consistent metadata for every model file; verifies downloaded model file hashes before installation; rejects non-HTTPS model downloads; marks OpenAI transcription models ready only when an API key is configured. src/main/ssh/connection/ssh-connection-sftp-namespace.test.ts: writes to the SFTP namespace path discovered on the same session; uploads the bundle into the SFTP namespace directory; issues no discovery requests when the caller supplies no mapping; ignores a mapping on a Windows host; never resolves on the system-SSH transport; leaves writeBuffer and openFileUploadSession on the shell path; keeps the shell path when discovery is inconclusive; preserves a session-limit channel-open error unchanged. src/main/ssh/connection/ssh-connection-system-transport.test.ts: uses system SSH transport when ProxyUseFdpass is resolved by OpenSSH; allows concurrent exec commands for system SSH with an Orca ControlMaster socket; keeps concurrent exec commands disabled for system SSH without a reusable socket; uses system SSH transport for ProxyCommand targets before ssh2 auth; keeps the original ssh2 reachability error when the system SSH probe fails; reports the system SSH failure when ssh2 cannot negotiate with the host; uses system transport for target or resolved OpenSSH proxy directives; allows an environment override for e2e coverage; ignores stale imported proxy fields when fresh OpenSSH config has no proxy. src/main/ssh/orcad-migration-catalog-client.test.ts: keeps each server project’s execution host, so a nested SSH row is never read as local; reads the server’s folder workspaces and the groups that place them on a host; sends and validates the %s operation; validates an abort result without letting it erase committed state; validates a committed receipt against the requested manifest; rejects malformed stage timestamps and abort envelopes; fails closed for method absence, malformed identity, and invalid pairing; refuses a destination without the capability before sending any migration request; reads method-not-found as the same refusal, not as a lost answer. src/main/ssh/runtime-ssh-access.test.ts: persists intent then claim before connecting and publishes redacted authenticated access; names the saved dependency when dependent state blocks an SSH access link; handles lost link responses without connecting or verifying again; invalidates old transport only after authenticated access is published; rejects changed request tuple after a completed link; rejects provisioning and missing durable generations without connecting; treats a lost successful unlink response as an idempotent no-op; hashes connection identity, not display state or owner; strictly validates ports and request IDs. src/main/startup/cli-launch-redirect.test.ts: detects a launch that received the unpacked CLI entrypoint; ignores normal desktop launches; ignores the entrypoint when it is only the executable itself (argv[0]); applies on Linux too; strips injected Chromium switches before node-mode CLI arguments; keeps user flags after the command and malformed boolean assignments; does not treat a later positional entrypoint path as the launcher; redirects a direct binary launch with no AppImage env at all; strips Chromium switches node mode would reject; preserves desktop-shaped switches after the command; leaves direct serve in-process but redirects its help; treats help as a CLI launch even without a command; treats %s as a CLI launch even without a command; does not treat a value of %s as a CLI early-exit flag; does not treat a serve option value as a help request; does not reinterpret help after the argument terminator; leaves a plain desktop launch alone; skips flag values when looking for the command positional; preserves a project selector in %j; does not apply the command form on macOS or Windows; never redirects an unpackaged build; runs the in-package CLI in Electron node mode with sanitized env; refuses to redirect twice so a dropped ELECTRON_RUN_AS_NODE cannot loop; reports a missing CLI entrypoint instead of booting the desktop app; execs a direct serve into the CLI so it keeps its pid and honours profile and supervise; keeps ORCA_USER_DATA_PATH for a direct serve without --user-data-dir; serves in-process when this runtime cannot exec or the exec fails; never execs the --serve child the CLI itself spawns, or serve on other platforms; surfaces a spawn failure as a non-zero exit; re-execs a direct --serve onto X11 when Electron resolved Wayland; leaves $name alone. src/main/startup/main-process-runtime-launch-activation.test.ts: marks startup settled for the launch record only after the window has loaded and settled; loads SSH targets on a windowless serve before paired clients can connect (#25886); pins the serve listener to loopback only for orcad browser sidecars (sidecar=$sidecar, #31895); saves Codex auth before a headless serve update restart, like the desktop does. src/main/startup/run-electron-vite-dev.test.ts: forwards dev instance identity to electron-vite; prepares userData orca and orca-dev wrappers for dev terminals; consumes the stable-name flag before forwarding args to electron-vite. src/main/updater/releases/updater-prerelease-feed.test.ts: returns the newest stable tag when the user is on an RC and a newer stable exists; returns the newest RC tag when no stable is newer than the current RC; can exclude prerelease tags for stable-channel checks; probes the %s %s platform manifest; returns null for stable-channel checks when only prereleases are newer; returns null when nothing in the feed is newer than the current version; ignores entries with unparseable tags; returns null when the fetch is not ok; returns null when the fetch throws; picks semver-newest across a mixed-order feed; ignores perf-tagged prereleases for regular RC checks; reports no RC update when only perf-tagged prereleases are newer; picks the semver-newest perf-tagged prerelease; matches only literal rc.N.perf prerelease tags for perf checks; reports no perf update instead of falling back to stable or RC tags; returns a bounded fallback candidate after the newest newer tag; returns null when the only newer tag has a manifest but its asset still 404s; requires a known old timestamp for a stable-fallback tag. src/main/updater/updater-events.test.ts: hands the whole downloaded event to artifact capture; publishes manual-install recovery after a %s download; publishes the normal downloaded state for AppImage builds; blocks downloaded-state handling when the packaged marker is unusable; passes the actual updater error into the install-failure handler; keeps manual-install recovery when a later check finds only the installed release; drops the artifact when another version takes over the cycle; ignores a downloaded event for an older target; accepts a matching event when the pending cache target was cleared; drops the artifact when progress reports a different pending version. src/main/updater/updater-linux-package-recovery-actions.test.ts: rejects both actions while no package-install recovery is active; validates the retained package on every invocation; restores recovery after a recheck resolves without a terminal event; restores recovery after a recheck fails; restores recovery when a pinned check resolves to the current version; restores recovery when resolving a pinned check fails; clears recovery for both actions once the package is gone; resolves %s as a result and keeps the card usable instead of rejecting; keeps recovery available after a transient read failure. src/main/updater/updater.check-settlement.test.ts: leaves a dismissed release update on the release source; surfaces net::ERR_FAILED to user-initiated checks with a friendly message; shows checking immediately for a user-initiated check while feed pinning is pending; keeps background checks event-driven before checking-for-update fires. src/main/usage/usage-provider-store-lifecycle.test.ts: persists identities for a whole first scan with one write; skips disabled and fresh matching states; invalidates prior sources on fingerprint changes and reuses them for forced scans. src/main/window/createMainWindow-close-confirmation.test.ts: clears the quit latch when the renderer prevents unload; does not notify the crash recorder when renderer teardown follows a confirmed window close; allows close when Electron reports a crashed webContents; requests confirmation for a hung-but-alive renderer instead of bypassing. src/main/workspaces/catalog/worktree-identity-persistence.test.ts: stores independent metadata for colliding locators on two hosts; projects canonical-only metadata for exactly one host; keeps canonical metadata current for legacy writers on a known owner; removes only the selected legacy-owner metadata when locators collide; keeps the canonical identity stable when the locator changes; leaves other hosts at the old locator when one host renames its folder. src/main/wsl/filesystem/wsl-drive-unc-conversion-startup.test.ts: leaves a journaled conversion untouched with the gate closed, %s. src/main/wsl/orcad/wsl-orcad-deploy.test.ts: holds the distro before it deploys and launches, and reports what C5 dials; passes its connect-state read (%s) on as hints; applies the SSH census and downgrade admission to the first deploy; refuses a WSL 1 distro before holding it or touching its files; tells an unregistered distro how to find the installed ones, before holding anything; surfaces an unsupported libc as the SSH host refusal does; starts an already-active slot the distro stop ended, instead of reporting it ready; deploys over an adopted unrecorded orcad under the SSH terminal gate, never a fresh launch; reports the adopted instance once the same version is serving; returns a deferral rather than launching over live terminals; hands a failed first deploy to the fresh-install cleanup before reporting it. src/main/zcode/zcode-v2-plan-credentials.test.ts: prefers migrated credentials over a stale legacy account; keeps legacy support when no personal v2 file exists; honors the upstream personal config override independently of the credential directory; does not borrow credentials from another execution host; preserves the protected Orca key priority even if v2 is corrupt; rejects upstream-invalid selection %j without legacy fallback; accepts a valid optional reasoning level; rejects upstream-invalid rule containers %j without legacy fallback; accepts a valid selected-provider empty personal rule and smart model rule; cannot establish selected credentials from %s rules; leaves encrypted CLI files unchanged across repeated account reads. tests/e2e/automation-terminal-launch.unit.test.ts: launches explicit Chat for %s through the common entry and keeps its actual fallback; does not apply stored Terminal extras to Chat; forwards only complete Chat-consumed values, with no Terminal startup base; refuses an expired admission without minting another launch identity; records the actual chat identity and delegates new workspace setup and provenance. tests/e2e/github-project-provider-columns.unit.test.ts: renders %s through one bounded items request on a named cloud host; preserves zero and complete counters: %j; preserves populated and legacy rows through shared sorting, grouping and cells. tests/e2e/structured-agent-session-launch-continuation.unit.test.ts: carries renderer lifetime through host intent to the real %s provider builders. tests/e2e/terminal-account-restart.unit.test.ts: replaces a tombstoned owner while its persisted pane binding has not yet been cleared; paired restart retains the old owner (host authority: %s). tests/e2e/terminal-hidden-view-parking.spec.ts: does not park excluded tabs; reproduces a static frame byte-for-byte across 25 park/reveal cycles. tests/e2e/terminal-long-table-scroll-restore.spec.ts: reproduces long markdown table artifacts after workspace switch and scroll; keeps narrow wrapped signer markdown table coherent after restore and scroll; keeps real emoji markdown table right edge clean after restore and scroll. tests/e2e/terminal-long-table-fixtures.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-paste-ownership.spec.ts: Windows multiline keyboard paste normalizes terminal newlines with one PTY owner; right-click paste sends clipboard text to the focused terminal exactly once. tests/e2e/terminal-raw-emoji-table-scroll-restore.spec.ts: uses WebGL by default for ordinary terminal output when available @terminal-rendering-golden. tests/e2e/terminal-windows-shell-paste-ownership.spec.ts: cmd.exe default terminal keyboard paste preserves exact content with one PTY owner; Git Bash default terminal keyboard paste preserves POSIX-shaped content with one PTY owner; WSL terminal keyboard paste preserves Linux shell content with one PTY owner. tests/e2e/workspace-board-lane-virtualization.spec.ts: mounts a window of cards for a large lane and keeps lane indexes; renders later lane indexes after the lane scrolls. Current changed source bytes, wrapper registrations and reduced selectors have no fresh execution or red/green credit. Unrelated surviving receipt objects keep their original narrow scope; current literal graph reachability does not establish historical import closure or equivalent replacement coverage. Current conservative command/import graph reaches src/main/runtime/orca-runtime-tests/terminal-sleep-and-teardown-part-03.spec.ts, src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-04.spec.ts, src/main/runtime/orca-runtime-tests/hooks-and-hosted-review-part-02.spec.ts, src/main/runtime/orca-runtime-tests/terminal-listing.spec.ts, src/main/runtime/orca-runtime-tests/window-authority.spec.ts, src/main/runtime/orchestration/mailbox/routing/orchestration-mailbox-routing-races.spec.ts, src/main/runtime/workspaces/removal/worktree-teardown-unstopped-pty.test.ts, src/main/skills/bundled-skill-install/layouts.test.ts, src/main/skills/skill-upload-session-service.test.ts, src/main/source-control/hosted-review-creation.test.ts, src/main/source-control/hosted-review-execution-host-routing.test.ts, src/main/source-control/hosted-review-inflight-invalidation.test.ts, src/main/speech/model-manager.test.ts, src/main/ssh/connection/ssh-connection-sftp-namespace.test.ts, src/main/ssh/connection/ssh-connection-system-transport.test.ts, src/main/ssh/orcad-migration-catalog-client.test.ts, src/main/ssh/runtime-ssh-access.test.ts, src/main/startup/cli-launch-redirect.test.ts, src/main/startup/main-process-runtime-launch-activation.test.ts, src/main/startup/run-electron-vite-dev.test.ts, src/main/updater/releases/updater-prerelease-feed.test.ts, src/main/updater/updater-events.test.ts, src/main/updater/updater-linux-package-recovery-actions.test.ts, src/main/updater/updater.check-settlement.test.ts, src/main/usage/usage-provider-store-lifecycle.test.ts, src/main/window/createMainWindow-close-confirmation.test.ts, src/main/workspaces/catalog/worktree-identity-persistence.test.ts, src/main/wsl/filesystem/wsl-drive-unc-conversion-startup.test.ts, src/main/wsl/orcad/wsl-orcad-deploy.test.ts, src/main/zcode/zcode-v2-plan-credentials.test.ts, tests/e2e/automation-terminal-launch.unit.test.ts, tests/e2e/github-project-provider-columns.unit.test.ts, tests/e2e/structured-agent-session-launch-continuation.unit.test.ts, tests/e2e/terminal-account-restart.unit.test.ts, tests/e2e/terminal-hidden-view-parking.spec.ts, tests/e2e/terminal-long-table-scroll-restore.spec.ts, tests/e2e/terminal-long-table-fixtures.ts, tests/e2e/terminal-paste-ownership.spec.ts, tests/e2e/terminal-raw-emoji-table-scroll-restore.spec.ts, tests/e2e/terminal-windows-shell-paste-ownership.spec.ts, tests/e2e/workspace-board-lane-virtualization.spec.ts. Unresolved local/computed edges remain explicit debt.", @@ -8224,13 +8225,13 @@ "providers": ["paired-runtime", "ssh"], "coveredPlatforms": ["macos"], "coveredProviders": [], - "coverageNotes": "Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. The decorative fanout suite retains only real OSC9999 ingestion through runtime/title heartbeat/viewer past the stale boundary, preserving stateStartedAt and clearing expired prompt metadata. Spinner volume/zero-work inventories, Pi/Grok frame tables, rich-status and foreground-reset examples are retired. Existing receipts for separate paired-terminal sources retain their original scopes. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The retained runtime registration driver loses three complete imported fragments: changed-working-label title facts and suppressed-frame tracked snapshots; refusing SSH GitLab MR-base lookup before local Git/glab calls; and missing folder-workspace path creation/update refusal before spawn/store mutation. These alarms are intentionally lost. Other imported fragments and separately rooted dated receipts are unchanged; they are not equivalent replacement coverage. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3342 removes three supplied activation/inventory fallback and same-handle retry examples; their fallback/retry alarms are intentionally lost without equivalent replacement. The exact mounted-HUB pending old-stream-end versus replacement-snapshot registration and its assertion remain. Unrelated dated macOS receipts retain only their original scope; no fresh runtime, provider, performance, mobile, mixed-version or cross-platform validation is claimed. NEXT46 terminal-owner layout and whole-case policy: src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-hidden-replay.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-visibility-and-exit.test.ts, src/renderer/src/runtime/hosted-terminals/streams/remote-runtime-terminal-stall-recovery.test.ts. Complete retired/reduced selectors lose their listed alarms; retained cases have no transferred historical pass or equivalent replacement coverage. Unaffected dated receipts retain only their original exact scopes. This metadata edit is not fresh execution evidence. Historical prior account (preserved verbatim): Audit3326 adopts reconciled source audit3324 and intentionally retires these low-value examples and their regression alarms: restarts only the stream whose renderer delivery credit never settles — An unreturned renderer delivery credit stalls only its stream, leaving healthy sibling/shared transport alive. Fake-clock deadline plus real multiplexer unsubscribe isolates recovery ownership.; probes then restarts a stream when an entered command receives no frames — Entered command produces no output; actual multiplexer must probe then close only its stalled stream after probe timeout. Distinct response liveness from parser credit.; establishes a probe baseline before recovering an unsequenced stream — 142-line manual five-probe baseline/regression sequence derives request IDs from encoded mock traffic and checks repeated no-close/close choreography. Oversized algorithm/state-machine example; no actual renderer/host command or recovery delivery. Retires unsequenced/unchanged/regressed baseline classification alarms.. Retained tests do not replace these alarms. Dated unrelated evidence remains scoped to its original execution; no historical source/import-closure equivalence, fresh execution, real network/provider/SSH, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Audit 3318 reduces remote-runtime-pty-transport-unpublished-host-graph.test.ts, retiring its direct published-empty-graph omission-to-pane-exit alarm. The unchanged bounded unpublished-frame/pushed-recovery and stale-handle park/manual-retry cases are narrower, not equivalent replacements. Both complete three-case Linux receipts dated 2026-11-04 are archived as historical, including source heads; they do not validate this reduced selector. Remaining fourteen active receipts are macOS only. No fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. The direct side-effect-facts selector retains six cases: replay exclusion, live sequence handling, stale disposer ownership, one handoff drain, drain-before-snapshot, and stale replacement-snapshot refusal. Successful current-entry snapshot application is not witnessed by the retained snapshot case. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2162: Reduced inventory-authority source intentionally loses the examples recorded in knownGaps; historical execution is not validation of changed bytes. Historical prior account (preserved verbatim): Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2090 (sources2088/2089): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2086 (sources2082/2083): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-handles-part-02.spec.ts: recovers a moved SSH pane through the tab it now sits in, not the one its lease froze; reports a plain expired orphan; does not report a pane whose only recent lease was superseded; does not report a pane whose only recent lease had its relay id recycled; picks the eligible orphan over a superseded predecessor that matches first; collects the same tabs the per-tab reader reports, in one sweep of the leases; reports no candidate when no lease names any of the worktree tabs. src/main/runtime/orca-runtime-tests/runtime-availability.spec.ts: starts unavailable with no authoritative window; reports the configured Windows terminal shell on status; reports floating workspace availability from settings on status; polls floating tabs with targeted PTY liveness and no repo/provider inventory; hydrates persisted tabs when the store cannot report repos; advertises browser screencast only when a renderer window is available; routes mobile Codex reset consumption through the account mutation coordinator; maps a definite pre-provider rejection into an authoritative current snapshot. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2003: Remove SSH/absent-inventory attach refusal, later-subscriber retry, no-view adoption/takeback, provider-tail/unknown-read projections, failed-keystroke retry and simpler loss reattach examples. Keep the existing-subscriber inventory-ready retry alias only at its retained callback, plus retained held-input cancellation/replacement and overlapping attach/loss fences. The inventory driver also loses the whole empty-worktree-publication registration and folder-host/unverifiable scalar examples. Retire the entire dated10-case attach receipt; retain16 independent receipts only in their original scopes. No retained-memory or real-daemon measurement follows from supplied provider inventories. Changed-source prior results do not validate reduced-source bytes. No fresh execution, equivalent replacement, red/green, measured budget, soak, physical provider/PTY, rendered, live remote/mobile or cross-platform result is claimed. Historical prior account (preserved verbatim): Audit1991: Remove initial-snapshot sequence-domain, limited-tail/truncation and unknown-output-sequence aliases whose callbacks are removed. The source retains partial-chunk replay and requested-snapshot retry/admission at supplied snapshot/data seams, not equivalent replacement or measured retained-memory proof. Preserve the existing stream/pending-wait capacity aliases only to their retained owning multiplex callbacks with supplied PTY resolutions. Removing the sibling-PTY heartbeat dedup work-count sample does not establish a new decorative fanout budget; the unrelated-PTY cancellation fence survives. Preserve17 unrelated dated receipts. Changed-source prior results do not validate reduced-source bytes. No fresh execution, equivalent replacement, red/green, measured budget, soak, physical provider/PTY, rendered, live remote/mobile or cross-platform result is claimed. Historical prior account (preserved verbatim): Audit1983: Remove the reasonless-close selector and narrow census aliases to actual retained runtime and injected-inventory callbacks. The removed synthetic census-dedup and isolated pre-boundary timing cases receive no replacement credit; preserve17 unrelated dated receipts. Changed-source prior results do not validate reduced-source bytes. No fresh execution, equivalent replacement, red/green, measured budget, soak, provider/PTY, rendered, live remote/mobile or cross-platform result is claimed. Historical prior account (preserved verbatim): Audit1939 (sources1934/1935): Delete the synthetic watcher-synchronization and partial-reconciliation suites. Retire all affected host-fact, capability-owner, activation-residue and retention-cap aliases because the remaining callbacks do not establish those full claims. Keep exact local hook/timer, byte-stream side effects, exit/disposal, scroll-intent cleanup and deterministic local rank/TTL progression callbacks within their mocked-host scope. No physical paired-host retention, resource budget or equivalent replacement evidence is assigned. The generic broad test command remains a runnable discovery command, never a dated execution receipt. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Delete the complete hook-agent-status projection suite and its two behavioral assertion aliases. Supplied stored hook/title/resume/staleness/rename/question/generation/PTY-retirement/snapshot-version projections are intentionally lost. The separate osc-status-scan gate mentions this source only in existing historical notes and receives no mutation. No measured hidden-output retention, CPU/RSS or resource budget result is claimed by the deleted source or this audit. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Module ownership cleanup retires the modeled cold-park private watcher/render/work-count suites under the current test policy. Their assertions are no longer active evidence. Retained lifecycle tests and headed/headless journeys do not establish equivalent coverage for those removed harness alarms. Audit1640 removes low-value IPC, SSH, workspace and transcript-watcher tests without equivalent replacement coverage. Per-source reviews record retained public refusal, cancellation, execution-host and real-filesystem outcomes with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1634 removes low-value IPC, filesystem and runtime projections, argument inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained cancellation, ownership, authorization, filesystem and SQLite boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1627 removes low-value mocked daemon projections, scalar tables and duplicate terminal examples without equivalent replacement coverage. Per-source reviews record retained terminal content preservation, process ownership refusal and recovery boundaries together with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Retained exact unchanged-source receipts remain the only current execution evidence. Reduced selectors have no fresh execution result. Audit1417 intentionally deletes src/renderer/src/components/terminal-pane/pty-connection-hidden-tui-snapshot-replay.spec.ts: Loses rich hidden snapshot sequence replay/grid resize and forced repaint after hidden snapshot alarms, including live-trigger deduplication. deletes src/renderer/src/components/terminal-pane/terminal-cold-park-verdict-loop.test.tsx: Loses modeled React oscillation alarm; real pane/watchers are not exercised. deletes src/renderer/src/runtime/remote-runtime-terminal-parse-backpressure.test.ts: Loses delayed credit until parsing, malformed frame/unknown stream handling, discarded transformed output recovery and failed ACK rearm alarms. No equivalent replacement coverage or fresh result is claimed. Historical coverage account: Audit1412 changes referenced sources. Retained unchanged-source receipts apply only to their unchanged commands; the reduced selectors have no new execution result. Historical coverage account (not a current claim for changed sources): This audit removes the browser response-after-manual-disconnect example. Its distinct in-flight response fencing alarm is lost; retained queued-call disconnect tests are not an equivalent replacement. Historical commands and results remain historical. Deterministic headed macOS runs launch an isolated Orca desktop server and a separate paired web client. A byte-identical headless run uses an isolated `orca serve` host. Both create six real paired host PTYs with bounded high-output scrollback, prove sustained output while all six client xterms are warm-mounted but hidden causes zero renderer scheduler work, then drive 180 timed real-PTY spinner frames through a second production `session.tabs.subscribeAll` transport and the real Zustand store. Each worktree must publish and apply only its exact initial-working and final-idle titles. Both topologies ordinary-park five xterms without enabling the lossy retention budget, assert bounded cells/heap/timer lag, then restore the bounded authoritative tail exactly once on the original PTY including output produced while parked and continued input/output. The headed oracle additionally keeps every remote target workspace and terminal unmounted in the host renderer before and after client recovery. Production-parity React contracts reproduce and remove the independent cold-park watcher/title feedback loop, including memo-cache discard, StrictMode cleanup replay, detached split-PTY replacement, and the zero-park fast path. A deterministic cross-boundary contract drives 24 real PTY trackers in distinct worktrees through 2,880 decorative spinner and bare-native redraw chunks and counts host publications, serialized bytes, renderer apply calls, and store mutations while preserving exact per-PTY raw chunks and per-worktree semantic transitions. Fake-clock continuation beyond the 30-minute stale boundary requires one deduplicated refresh per worktree every 15 minutes, globally spaced by at least 50 ms, with every mirrored working status still fresh. Exact-identical Pi plus rotating normalized Pi/Grok frames renew coarse working/permission state; rich prompt/tool/question fields retain their original 30-minute lease, and `stateStartedAt`/acknowledgement identity remain stable across title/hook source swaps. A separate headed desktop-client journey relaunches the host main process while retaining the daemon and a daemon-backgrounded PTY, then requires first-observation foreground sync and continued input/output on the same PTY. Unit coverage keeps layout versions out of snapshot output high-waters when a restored snapshot has no output sequence and retries an existing subscriber only after authoritative provider inventory proves the retained PTY exists. Separate headed/headless ACK-starvation tests recover one stalled stream without replacing its PTY. A headed output-drop oracle proves a successful snapshot probe cannot certify a stale live stream when the authoritative PTY sequence advanced beyond the client high-water. Unit coverage requires an unsequenced client to establish a first-probe baseline, remain attached at the same sequence, and recover only after a later probe advances. A capability-disabled run proves legacy hosts retain the prior lossy limit/TTL fallback. Unit tests cover exact-owner capability routing, raw-stream release, singleton side-effect facts with timed handoff cleanup, 64 simultaneous synthetic-title sources with zero decorative client events after convergence, per-client late-subscribe recovery, semantic bell/title transitions, local animation-frame preservation, legacy no-output-pause delivery gating with exact snapshot restore, provider-authoritative snapshots, 128 active streams plus retry after a 129th-stream capacity rejection, capacity-pressure backoff, full split-leaf remint reconciliation, truncation, and manual-disconnect queue fencing. Linux, Windows, live SSH, and production-scale paired hosts remain gaps. The audit removed the E2E parking override/debug-handle suite. Override exposure guards, delay/retention independence, TTL filtering and debug-handle forwarding alarms are lost. Retained real parking journeys do not establish equivalent test-harness configuration coverage. The test-value audit removes separate responsive-stream, intervening-control-frame and capacity-error classification examples. Retained multiplexer checks use fake clocks and captured transport callbacks; their per-stream isolation, deadline and probe-baseline assertions do not prove an actual host restart or equivalent replacement coverage. Historical run commands and results remain historical. The test-value audit removes the standalone retained/live metadata projection and additive-preference negative example from runtime-client-events.test.ts. Its remaining case checks partial-authority event admission through a captured callback; it does not establish full-pair projection, actual peer delivery or equivalent replacement coverage. Historical commands and results remain historical. The isolated accepted-handle-events observer suite is removed. Its runtime/worktree/pane membership, once-only delivery and listener-release ledger alarms are lost. Live hidden-retention evidence does not establish equivalent observer coverage. Audit1292: The entire host-session mirror empty-inventory settle suite is removed. Live-unpublished versus genuinely-empty hydration, same-ID revision response/in-flight fencing and cleared-tracking pending-probe invalidation alarms are lost. The singleton store and inventories were mocked; no real host census or resume delivery or equivalent replacement coverage is claimed. Audit1302: src/main/runtime/session-tabs-inventory-publication.test.ts: Keep actual runtime public inventory authority across renderer epochs/resync, fallback, aborts and invalid graph publishers. Remove copied nonempty publication, private scan counters/collector and known-host replacement, fake refresh that performs the restoration itself, environment capability flag, and private resume-store characterization whose fixture supplies reservation/rollback ordering. Lose isolated refresh/churn/receiver/capability and copied-payload wiring alarms; no equivalent replacement. Reduced selectors require fresh validation; prior full-suite receipts do not validate removed or changed coverage. No equivalent replacement or physical-provider claim. Audit1312 removes low-value imported specs from aggregate test drivers; old whole-driver execution receipts are archived. Retained registrations are unchanged. No equivalent replacement coverage is claimed. Audit1318: src/shared/remote-runtime/control/remote-runtime-shared-control-connection.test.ts: Retain real local encrypted WebSocket regressions for pre-ready socket replacement preserving a request, close-before-ready logical subscription cleanup, unrelated RPC surviving another deadline, and pending setup abort leaving a sibling pending. Host server is synthetic and uses production crypto; kept scenarios also retain internal diagnostics/memory assertions unchanged. Remove happy routing, authority-envelope, unknown-ID/log, cleanup enum, binary, keepalive/liveness and production-budget/private socket inventories; intentionally lose those alarms, no physical host authorization, independent crypto or remote provider claim. Reduced selectors need fresh validation; historical full-suite receipts do not validate the changed coverage. No equivalent replacement or physical-provider claim. Audit1346 removes the entire private-state-injected session-tabs rich-status boundary suite. Its stale task/question, neutral-title preservation, owner-change and provider-generation alarms are intentionally lost. Retained real terminal journeys do not establish equivalent replacement coverage. The prior mixed 42-contract execution receipt is historical and must not validate the reduced current selector; fresh validation remains required. Audit1352: focused Linux unit selections validated on source head b638e07d395b76af3ba6b03761a1a0fe7e5e026e. Historical statements above retain their original scope. Intentional test losses are recorded in knownGaps; these runs claim no physical-provider, p95, rendered E2E or cross-platform validation. Audit1360: focused Linux unit selections validated on source head 415c9cffc2d10024131ec4b6721767a22195de6d. Historical statements above retain their original scope. Intentional test losses are recorded in knownGaps; these runs claim no physical-provider, p95, rendered E2E or cross-platform validation. Audit1383: src/main/runtime/orca-runtime-tests/worktree-native-chat-setup.spec.ts reduced. Lose command-list/receipt/default/background activation matrices whose provider spawn returns canned IDs. Retained test proves actual tab allocation separation, not terminal startup. No equivalent replacement coverage is claimed. Existing exact unaffected receipts and their platform/provider/budget claims remain scoped to unchanged selectors. Changed optional selectors have no new execution result. Audit1391: src/renderer/src/components/terminal-pane/pty-connection-remote-pause-drop-notice.spec.ts reduced. Lose repeated quiet switches avoiding snapshots and legacy host without drop-report capability requesting a snapshot. Retained imported registrar uses fake transport/xterm writes, exercising real gap/restore lifecycle rather than actual terminal rendering. No equivalent replacement coverage is claimed. The existing primary-selector receipt and its claims remain scoped to their unchanged sources; the broad optional command has no new execution result. Audit1396: src/renderer/src/components/terminal-pane/terminal-parked-watcher-reconciliation.test.ts reduced. Retire standalone reconcile set output and positive untouchedFreshSpawn/workspaceOwner copying alarms; no native watcher execution claim. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result. Audit1401: src/main/runtime/orca-runtime-tests/mobile-session-tabs-duplicate-pty-repair.spec.ts removed. Loses listing-on-already-repaired-data alarm; it never established runtime repair invocation. src/main/runtime/orca-runtime-tests/mobile-session-tabs-split-agent-tab.spec.ts removed. Loses actual mobile split sibling shell versus recorded-agent resume command alarms; no equivalent coverage asserted. src/main/runtime/orca-runtime-tests/worktree-native-chat-setup.spec.ts removed. Loses separate setup/startup tab result in that manually assembled path. src/main/runtime/orca-runtime.test.ts reduced. Driver registration/import of a whole-deleted selected spec is removed; directly authored cases remain unchanged. No extra authored removal credit. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result. Audit1402: src/renderer/src/components/terminal-pane/pty-connection-park-reveal-model-snapshot.spec.ts reduced. Loses basic paired serialize/initial replay forwarding check; hidden alt-frame grid/precedence policy remains. No real terminal renderer claim. No equivalent replacement coverage is claimed. The existing primary-selector receipt and its claims remain scoped to their unchanged sources; the broad optional command has no new execution result. Audit1412: src/main/runtime/mobile-session-tabs-agent-status-heartbeat.test.ts reduced. Loses FIFO fairness/spacing, semantic bypass/cancellation, delayed-boundary scheduling and teardown timer-cleanup alarms. Retained fixed emission cases do not replace those outcomes. No equivalent replacement coverage is claimed.", + "coverageNotes": "Audit3563 scope: retire the whole Kiro OSC/hook identity suite and the OpenCode2 identity-only case; reduce Codex cached startup child-retirement and missing-session-ID tables to one case each (2 authored/4 finite removals). Intentionally lose Kiro alias/session continuity, OpenCode2 completed OSC identity continuity, cached replacement without relay-generation restart, and missing-previous-ID context-preservation alarms. Remaining provider, remote compatibility, persistence and cross-session pollution cases are not equivalent replacements. Also remove the private Codex roster-size assertion while retaining public child-absence checks. Historical evidence retains its original source scope; no new execution or transferred pass validates these reductions. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. The decorative fanout suite retains only real OSC9999 ingestion through runtime/title heartbeat/viewer past the stale boundary, preserving stateStartedAt and clearing expired prompt metadata. Spinner volume/zero-work inventories, Pi/Grok frame tables, rich-status and foreground-reset examples are retired. Existing receipts for separate paired-terminal sources retain their original scopes. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The retained runtime registration driver loses three complete imported fragments: changed-working-label title facts and suppressed-frame tracked snapshots; refusing SSH GitLab MR-base lookup before local Git/glab calls; and missing folder-workspace path creation/update refusal before spawn/store mutation. These alarms are intentionally lost. Other imported fragments and separately rooted dated receipts are unchanged; they are not equivalent replacement coverage. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3342 removes three supplied activation/inventory fallback and same-handle retry examples; their fallback/retry alarms are intentionally lost without equivalent replacement. The exact mounted-HUB pending old-stream-end versus replacement-snapshot registration and its assertion remain. Unrelated dated macOS receipts retain only their original scope; no fresh runtime, provider, performance, mobile, mixed-version or cross-platform validation is claimed. NEXT46 terminal-owner layout and whole-case policy: src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-hidden-replay.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-visibility-and-exit.test.ts, src/renderer/src/runtime/hosted-terminals/streams/remote-runtime-terminal-stall-recovery.test.ts. Complete retired/reduced selectors lose their listed alarms; retained cases have no transferred historical pass or equivalent replacement coverage. Unaffected dated receipts retain only their original exact scopes. This metadata edit is not fresh execution evidence. Historical prior account (preserved verbatim): Audit3326 adopts reconciled source audit3324 and intentionally retires these low-value examples and their regression alarms: restarts only the stream whose renderer delivery credit never settles — An unreturned renderer delivery credit stalls only its stream, leaving healthy sibling/shared transport alive. Fake-clock deadline plus real multiplexer unsubscribe isolates recovery ownership.; probes then restarts a stream when an entered command receives no frames — Entered command produces no output; actual multiplexer must probe then close only its stalled stream after probe timeout. Distinct response liveness from parser credit.; establishes a probe baseline before recovering an unsequenced stream — 142-line manual five-probe baseline/regression sequence derives request IDs from encoded mock traffic and checks repeated no-close/close choreography. Oversized algorithm/state-machine example; no actual renderer/host command or recovery delivery. Retires unsequenced/unchanged/regressed baseline classification alarms.. Retained tests do not replace these alarms. Dated unrelated evidence remains scoped to its original execution; no historical source/import-closure equivalence, fresh execution, real network/provider/SSH, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Audit 3318 reduces remote-runtime-pty-transport-unpublished-host-graph.test.ts, retiring its direct published-empty-graph omission-to-pane-exit alarm. The unchanged bounded unpublished-frame/pushed-recovery and stale-handle park/manual-retry cases are narrower, not equivalent replacements. Both complete three-case Linux receipts dated 2026-11-04 are archived as historical, including source heads; they do not validate this reduced selector. Remaining fourteen active receipts are macOS only. No fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. The direct side-effect-facts selector retains six cases: replay exclusion, live sequence handling, stale disposer ownership, one handoff drain, drain-before-snapshot, and stale replacement-snapshot refusal. Successful current-entry snapshot application is not witnessed by the retained snapshot case. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2162: Reduced inventory-authority source intentionally loses the examples recorded in knownGaps; historical execution is not validation of changed bytes. Historical prior account (preserved verbatim): Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2090 (sources2088/2089): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2086 (sources2082/2083): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-handles-part-02.spec.ts: recovers a moved SSH pane through the tab it now sits in, not the one its lease froze; reports a plain expired orphan; does not report a pane whose only recent lease was superseded; does not report a pane whose only recent lease had its relay id recycled; picks the eligible orphan over a superseded predecessor that matches first; collects the same tabs the per-tab reader reports, in one sweep of the leases; reports no candidate when no lease names any of the worktree tabs. src/main/runtime/orca-runtime-tests/runtime-availability.spec.ts: starts unavailable with no authoritative window; reports the configured Windows terminal shell on status; reports floating workspace availability from settings on status; polls floating tabs with targeted PTY liveness and no repo/provider inventory; hydrates persisted tabs when the store cannot report repos; advertises browser screencast only when a renderer window is available; routes mobile Codex reset consumption through the account mutation coordinator; maps a definite pre-provider rejection into an authoritative current snapshot. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2003: Remove SSH/absent-inventory attach refusal, later-subscriber retry, no-view adoption/takeback, provider-tail/unknown-read projections, failed-keystroke retry and simpler loss reattach examples. Keep the existing-subscriber inventory-ready retry alias only at its retained callback, plus retained held-input cancellation/replacement and overlapping attach/loss fences. The inventory driver also loses the whole empty-worktree-publication registration and folder-host/unverifiable scalar examples. Retire the entire dated10-case attach receipt; retain16 independent receipts only in their original scopes. No retained-memory or real-daemon measurement follows from supplied provider inventories. Changed-source prior results do not validate reduced-source bytes. No fresh execution, equivalent replacement, red/green, measured budget, soak, physical provider/PTY, rendered, live remote/mobile or cross-platform result is claimed. Historical prior account (preserved verbatim): Audit1991: Remove initial-snapshot sequence-domain, limited-tail/truncation and unknown-output-sequence aliases whose callbacks are removed. The source retains partial-chunk replay and requested-snapshot retry/admission at supplied snapshot/data seams, not equivalent replacement or measured retained-memory proof. Preserve the existing stream/pending-wait capacity aliases only to their retained owning multiplex callbacks with supplied PTY resolutions. Removing the sibling-PTY heartbeat dedup work-count sample does not establish a new decorative fanout budget; the unrelated-PTY cancellation fence survives. Preserve17 unrelated dated receipts. Changed-source prior results do not validate reduced-source bytes. No fresh execution, equivalent replacement, red/green, measured budget, soak, physical provider/PTY, rendered, live remote/mobile or cross-platform result is claimed. Historical prior account (preserved verbatim): Audit1983: Remove the reasonless-close selector and narrow census aliases to actual retained runtime and injected-inventory callbacks. The removed synthetic census-dedup and isolated pre-boundary timing cases receive no replacement credit; preserve17 unrelated dated receipts. Changed-source prior results do not validate reduced-source bytes. No fresh execution, equivalent replacement, red/green, measured budget, soak, provider/PTY, rendered, live remote/mobile or cross-platform result is claimed. Historical prior account (preserved verbatim): Audit1939 (sources1934/1935): Delete the synthetic watcher-synchronization and partial-reconciliation suites. Retire all affected host-fact, capability-owner, activation-residue and retention-cap aliases because the remaining callbacks do not establish those full claims. Keep exact local hook/timer, byte-stream side effects, exit/disposal, scroll-intent cleanup and deterministic local rank/TTL progression callbacks within their mocked-host scope. No physical paired-host retention, resource budget or equivalent replacement evidence is assigned. The generic broad test command remains a runnable discovery command, never a dated execution receipt. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Delete the complete hook-agent-status projection suite and its two behavioral assertion aliases. Supplied stored hook/title/resume/staleness/rename/question/generation/PTY-retirement/snapshot-version projections are intentionally lost. The separate osc-status-scan gate mentions this source only in existing historical notes and receives no mutation. No measured hidden-output retention, CPU/RSS or resource budget result is claimed by the deleted source or this audit. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Module ownership cleanup retires the modeled cold-park private watcher/render/work-count suites under the current test policy. Their assertions are no longer active evidence. Retained lifecycle tests and headed/headless journeys do not establish equivalent coverage for those removed harness alarms. Audit1640 removes low-value IPC, SSH, workspace and transcript-watcher tests without equivalent replacement coverage. Per-source reviews record retained public refusal, cancellation, execution-host and real-filesystem outcomes with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1634 removes low-value IPC, filesystem and runtime projections, argument inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained cancellation, ownership, authorization, filesystem and SQLite boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1627 removes low-value mocked daemon projections, scalar tables and duplicate terminal examples without equivalent replacement coverage. Per-source reviews record retained terminal content preservation, process ownership refusal and recovery boundaries together with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Retained exact unchanged-source receipts remain the only current execution evidence. Reduced selectors have no fresh execution result. Audit1417 intentionally deletes src/renderer/src/components/terminal-pane/pty-connection-hidden-tui-snapshot-replay.spec.ts: Loses rich hidden snapshot sequence replay/grid resize and forced repaint after hidden snapshot alarms, including live-trigger deduplication. deletes src/renderer/src/components/terminal-pane/terminal-cold-park-verdict-loop.test.tsx: Loses modeled React oscillation alarm; real pane/watchers are not exercised. deletes src/renderer/src/runtime/remote-runtime-terminal-parse-backpressure.test.ts: Loses delayed credit until parsing, malformed frame/unknown stream handling, discarded transformed output recovery and failed ACK rearm alarms. No equivalent replacement coverage or fresh result is claimed. Historical coverage account: Audit1412 changes referenced sources. Retained unchanged-source receipts apply only to their unchanged commands; the reduced selectors have no new execution result. Historical coverage account (not a current claim for changed sources): This audit removes the browser response-after-manual-disconnect example. Its distinct in-flight response fencing alarm is lost; retained queued-call disconnect tests are not an equivalent replacement. Historical commands and results remain historical. Deterministic headed macOS runs launch an isolated Orca desktop server and a separate paired web client. A byte-identical headless run uses an isolated `orca serve` host. Both create six real paired host PTYs with bounded high-output scrollback, prove sustained output while all six client xterms are warm-mounted but hidden causes zero renderer scheduler work, then drive 180 timed real-PTY spinner frames through a second production `session.tabs.subscribeAll` transport and the real Zustand store. Each worktree must publish and apply only its exact initial-working and final-idle titles. Both topologies ordinary-park five xterms without enabling the lossy retention budget, assert bounded cells/heap/timer lag, then restore the bounded authoritative tail exactly once on the original PTY including output produced while parked and continued input/output. The headed oracle additionally keeps every remote target workspace and terminal unmounted in the host renderer before and after client recovery. Production-parity React contracts reproduce and remove the independent cold-park watcher/title feedback loop, including memo-cache discard, StrictMode cleanup replay, detached split-PTY replacement, and the zero-park fast path. A deterministic cross-boundary contract drives 24 real PTY trackers in distinct worktrees through 2,880 decorative spinner and bare-native redraw chunks and counts host publications, serialized bytes, renderer apply calls, and store mutations while preserving exact per-PTY raw chunks and per-worktree semantic transitions. Fake-clock continuation beyond the 30-minute stale boundary requires one deduplicated refresh per worktree every 15 minutes, globally spaced by at least 50 ms, with every mirrored working status still fresh. Exact-identical Pi plus rotating normalized Pi/Grok frames renew coarse working/permission state; rich prompt/tool/question fields retain their original 30-minute lease, and `stateStartedAt`/acknowledgement identity remain stable across title/hook source swaps. A separate headed desktop-client journey relaunches the host main process while retaining the daemon and a daemon-backgrounded PTY, then requires first-observation foreground sync and continued input/output on the same PTY. Unit coverage keeps layout versions out of snapshot output high-waters when a restored snapshot has no output sequence and retries an existing subscriber only after authoritative provider inventory proves the retained PTY exists. Separate headed/headless ACK-starvation tests recover one stalled stream without replacing its PTY. A headed output-drop oracle proves a successful snapshot probe cannot certify a stale live stream when the authoritative PTY sequence advanced beyond the client high-water. Unit coverage requires an unsequenced client to establish a first-probe baseline, remain attached at the same sequence, and recover only after a later probe advances. A capability-disabled run proves legacy hosts retain the prior lossy limit/TTL fallback. Unit tests cover exact-owner capability routing, raw-stream release, singleton side-effect facts with timed handoff cleanup, 64 simultaneous synthetic-title sources with zero decorative client events after convergence, per-client late-subscribe recovery, semantic bell/title transitions, local animation-frame preservation, legacy no-output-pause delivery gating with exact snapshot restore, provider-authoritative snapshots, 128 active streams plus retry after a 129th-stream capacity rejection, capacity-pressure backoff, full split-leaf remint reconciliation, truncation, and manual-disconnect queue fencing. Linux, Windows, live SSH, and production-scale paired hosts remain gaps. The audit removed the E2E parking override/debug-handle suite. Override exposure guards, delay/retention independence, TTL filtering and debug-handle forwarding alarms are lost. Retained real parking journeys do not establish equivalent test-harness configuration coverage. The test-value audit removes separate responsive-stream, intervening-control-frame and capacity-error classification examples. Retained multiplexer checks use fake clocks and captured transport callbacks; their per-stream isolation, deadline and probe-baseline assertions do not prove an actual host restart or equivalent replacement coverage. Historical run commands and results remain historical. The test-value audit removes the standalone retained/live metadata projection and additive-preference negative example from runtime-client-events.test.ts. Its remaining case checks partial-authority event admission through a captured callback; it does not establish full-pair projection, actual peer delivery or equivalent replacement coverage. Historical commands and results remain historical. The isolated accepted-handle-events observer suite is removed. Its runtime/worktree/pane membership, once-only delivery and listener-release ledger alarms are lost. Live hidden-retention evidence does not establish equivalent observer coverage. Audit1292: The entire host-session mirror empty-inventory settle suite is removed. Live-unpublished versus genuinely-empty hydration, same-ID revision response/in-flight fencing and cleared-tracking pending-probe invalidation alarms are lost. The singleton store and inventories were mocked; no real host census or resume delivery or equivalent replacement coverage is claimed. Audit1302: src/main/runtime/session-tabs-inventory-publication.test.ts: Keep actual runtime public inventory authority across renderer epochs/resync, fallback, aborts and invalid graph publishers. Remove copied nonempty publication, private scan counters/collector and known-host replacement, fake refresh that performs the restoration itself, environment capability flag, and private resume-store characterization whose fixture supplies reservation/rollback ordering. Lose isolated refresh/churn/receiver/capability and copied-payload wiring alarms; no equivalent replacement. Reduced selectors require fresh validation; prior full-suite receipts do not validate removed or changed coverage. No equivalent replacement or physical-provider claim. Audit1312 removes low-value imported specs from aggregate test drivers; old whole-driver execution receipts are archived. Retained registrations are unchanged. No equivalent replacement coverage is claimed. Audit1318: src/shared/remote-runtime/control/remote-runtime-shared-control-connection.test.ts: Retain real local encrypted WebSocket regressions for pre-ready socket replacement preserving a request, close-before-ready logical subscription cleanup, unrelated RPC surviving another deadline, and pending setup abort leaving a sibling pending. Host server is synthetic and uses production crypto; kept scenarios also retain internal diagnostics/memory assertions unchanged. Remove happy routing, authority-envelope, unknown-ID/log, cleanup enum, binary, keepalive/liveness and production-budget/private socket inventories; intentionally lose those alarms, no physical host authorization, independent crypto or remote provider claim. Reduced selectors need fresh validation; historical full-suite receipts do not validate the changed coverage. No equivalent replacement or physical-provider claim. Audit1346 removes the entire private-state-injected session-tabs rich-status boundary suite. Its stale task/question, neutral-title preservation, owner-change and provider-generation alarms are intentionally lost. Retained real terminal journeys do not establish equivalent replacement coverage. The prior mixed 42-contract execution receipt is historical and must not validate the reduced current selector; fresh validation remains required. Audit1352: focused Linux unit selections validated on source head b638e07d395b76af3ba6b03761a1a0fe7e5e026e. Historical statements above retain their original scope. Intentional test losses are recorded in knownGaps; these runs claim no physical-provider, p95, rendered E2E or cross-platform validation. Audit1360: focused Linux unit selections validated on source head 415c9cffc2d10024131ec4b6721767a22195de6d. Historical statements above retain their original scope. Intentional test losses are recorded in knownGaps; these runs claim no physical-provider, p95, rendered E2E or cross-platform validation. Audit1383: src/main/runtime/orca-runtime-tests/worktree-native-chat-setup.spec.ts reduced. Lose command-list/receipt/default/background activation matrices whose provider spawn returns canned IDs. Retained test proves actual tab allocation separation, not terminal startup. No equivalent replacement coverage is claimed. Existing exact unaffected receipts and their platform/provider/budget claims remain scoped to unchanged selectors. Changed optional selectors have no new execution result. Audit1391: src/renderer/src/components/terminal-pane/pty-connection-remote-pause-drop-notice.spec.ts reduced. Lose repeated quiet switches avoiding snapshots and legacy host without drop-report capability requesting a snapshot. Retained imported registrar uses fake transport/xterm writes, exercising real gap/restore lifecycle rather than actual terminal rendering. No equivalent replacement coverage is claimed. The existing primary-selector receipt and its claims remain scoped to their unchanged sources; the broad optional command has no new execution result. Audit1396: src/renderer/src/components/terminal-pane/terminal-parked-watcher-reconciliation.test.ts reduced. Retire standalone reconcile set output and positive untouchedFreshSpawn/workspaceOwner copying alarms; no native watcher execution claim. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result. Audit1401: src/main/runtime/orca-runtime-tests/mobile-session-tabs-duplicate-pty-repair.spec.ts removed. Loses listing-on-already-repaired-data alarm; it never established runtime repair invocation. src/main/runtime/orca-runtime-tests/mobile-session-tabs-split-agent-tab.spec.ts removed. Loses actual mobile split sibling shell versus recorded-agent resume command alarms; no equivalent coverage asserted. src/main/runtime/orca-runtime-tests/worktree-native-chat-setup.spec.ts removed. Loses separate setup/startup tab result in that manually assembled path. src/main/runtime/orca-runtime.test.ts reduced. Driver registration/import of a whole-deleted selected spec is removed; directly authored cases remain unchanged. No extra authored removal credit. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result. Audit1402: src/renderer/src/components/terminal-pane/pty-connection-park-reveal-model-snapshot.spec.ts reduced. Loses basic paired serialize/initial replay forwarding check; hidden alt-frame grid/precedence policy remains. No real terminal renderer claim. No equivalent replacement coverage is claimed. The existing primary-selector receipt and its claims remain scoped to their unchanged sources; the broad optional command has no new execution result. Audit1412: src/main/runtime/mobile-session-tabs-agent-status-heartbeat.test.ts reduced. Loses FIFO fairness/spacing, semantic bypass/cancellation, delayed-boundary scheduling and teardown timer-cleanup alarms. Retained fixed emission cases do not replace those outcomes. No equivalent replacement coverage is claimed.", "motivatingLinks": [ "https://github.com/stablyai/orca/issues/8652", "https://github.com/stablyai/orca/pull/10625" ], "invariant": "A host advertising terminal.paired-parking.v1 keeps the PTY and bounded authoritative history alive while an ordinary hidden-view park destroys the client xterm and releases its raw per-PTY stream. Reveal must restore up to the requested 5,000 rows, parked-time side effects/output, the same PTY identity, and continued input/output. Parked watcher synchronization compares parking semantics rather than fresh object identity, owns setup and cleanup across StrictMode replay, reconciles real PTY/layout changes, and performs no terminal-state scan when nothing is parked. After a host main-process relaunch, unknown local delivery-sync state must not be treated as proof that a surviving daemon PTY is already foregrounded. An empty headed session-tab inventory is authoritative only after the current renderer graph generation publishes a complete inventory and every execution host answers the PTY census; unpublished, resync-pending, transport-failed, and partially scoped inventories remain unverifiable, while genuinely authoritative empty and non-empty inventories settle and a scope-less legacy empty keeps its pre-authority best-effort settle so outdated hosts retain agent auto-resume. The unpublished-empty regression must keep the resume dispatch parked with exactly one daemon process/writer; restoring unconditional empty hydration must release one dispatch and create two live writers. If subscription precedes provider readiness, authoritative inventory must reconsider that existing subscriber without spawning, resizing, or changing reconnect state. A snapshot without an output high-water must keep it unknown rather than borrowing a layout version that can suppress newer live output. Hosts without the capability must gate hidden raw output before xterm scheduling and repaint from the authoritative snapshot on reveal while retaining the existing limit/TTL force-parking fallback. After the first semantic title state, decorative spinner frequency must add zero paired client-event frames and zero full session-tabs work between status-freshness leases regardless of hidden worktree count. Continuous working or permission evidence may refresh each affected worktree once per 15 minutes through one globally spaced FIFO so current and legacy viewers do not decay before 30 minutes; sibling PTYs share that worktree refresh, while local title animation, raw output, and semantic title, status, bell, completion, and query facts remain intact. A paired terminal stream whose delivery credits stop progressing must replace only that stream; command silence first probes authoritative state and replaces the stream if the probe times out or proves the PTY advanced beyond the client's delivered output sequence. When no comparable delivery high-water exists, the first sequenced probe establishes a baseline and only later advancement proves staleness. A same-sequence snapshot remains valid proof of a responsive silent command. A successful status probe may replace a pre-ready shared-control socket without rejecting or duplicating calls already waiting for that transport. Manual disconnect must retain pairing while preventing queued or passive calls and subscriptions from recreating transport until explicit Connect. A paired pane reconnecting to a relaunched host that has not yet published its window graph (an unpublished frame, bare or client-projected) must stay in recovery rather than retire, and the host's first published snapshot carrying the surface must reattach it; a published frame lacking the surface still retires it. The placeholder is only ever sent before the host's graph first publishes; afterwards a worktree with no tabs gets a real empty answer, pushed once to any client that asked during startup, so a paired client opening a never-opened host worktree still gets its first terminal.", - "oracle": "Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. The decorative fanout suite retains only real OSC9999 ingestion through runtime/title heartbeat/viewer past the stale boundary, preserving stateStartedAt and clearing expired prompt metadata. Spinner volume/zero-work inventories, Pi/Grok frame tables, rich-status and foreground-reset examples are retired. Existing receipts for separate paired-terminal sources retain their original scopes. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The retained runtime registration driver loses three complete imported fragments: changed-working-label title facts and suppressed-frame tracked snapshots; refusing SSH GitLab MR-base lookup before local Git/glab calls; and missing folder-workspace path creation/update refusal before spawn/store mutation. These alarms are intentionally lost. Other imported fragments and separately rooted dated receipts are unchanged; they are not equivalent replacement coverage. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3342 removes three supplied activation/inventory fallback and same-handle retry examples; their fallback/retry alarms are intentionally lost without equivalent replacement. The exact mounted-HUB pending old-stream-end versus replacement-snapshot registration and its assertion remain. Unrelated dated macOS receipts retain only their original scope; no fresh runtime, provider, performance, mobile, mixed-version or cross-platform validation is claimed. Audit3326 adopts reconciled source audit3324 and intentionally retires these low-value examples and their regression alarms: restarts only the stream whose renderer delivery credit never settles — An unreturned renderer delivery credit stalls only its stream, leaving healthy sibling/shared transport alive. Fake-clock deadline plus real multiplexer unsubscribe isolates recovery ownership.; probes then restarts a stream when an entered command receives no frames — Entered command produces no output; actual multiplexer must probe then close only its stalled stream after probe timeout. Distinct response liveness from parser credit.; establishes a probe baseline before recovering an unsequenced stream — 142-line manual five-probe baseline/regression sequence derives request IDs from encoded mock traffic and checks repeated no-close/close choreography. Oversized algorithm/state-machine example; no actual renderer/host command or recovery delivery. Retires unsequenced/unchanged/regressed baseline classification alarms.. Retained tests do not replace these alarms. Dated unrelated evidence remains scoped to its original execution; no historical source/import-closure equivalence, fresh execution, real network/provider/SSH, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Audit 3318 reduces remote-runtime-pty-transport-unpublished-host-graph.test.ts, retiring its direct published-empty-graph omission-to-pane-exit alarm. The unchanged bounded unpublished-frame/pushed-recovery and stale-handle park/manual-retry cases are narrower, not equivalent replacements. Both complete three-case Linux receipts dated 2026-11-04 are archived as historical, including source heads; they do not validate this reduced selector. Remaining fourteen active receipts are macOS only. No fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. The direct side-effect-facts selector retains six cases: replay exclusion, live sequence handling, stale disposer ownership, one handoff drain, drain-before-snapshot, and stale replacement-snapshot refusal. Successful current-entry snapshot application is not witnessed by the retained snapshot case. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2090 (sources2088/2089): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2086 (sources2082/2083): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-handles-part-02.spec.ts: recovers a moved SSH pane through the tab it now sits in, not the one its lease froze; reports a plain expired orphan; does not report a pane whose only recent lease was superseded; does not report a pane whose only recent lease had its relay id recycled; picks the eligible orphan over a superseded predecessor that matches first; collects the same tabs the per-tab reader reports, in one sweep of the leases; reports no candidate when no lease names any of the worktree tabs. src/main/runtime/orca-runtime-tests/runtime-availability.spec.ts: starts unavailable with no authoritative window; reports the configured Windows terminal shell on status; reports floating workspace availability from settings on status; polls floating tabs with targeted PTY liveness and no repo/provider inventory; hydrates persisted tabs when the store cannot report repos; advertises browser screencast only when a renderer window is available; routes mobile Codex reset consumption through the account mutation coordinator; maps a definite pre-provider rejection into an authoritative current snapshot. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2003: Remove SSH/absent-inventory attach refusal, later-subscriber retry, no-view adoption/takeback, provider-tail/unknown-read projections, failed-keystroke retry and simpler loss reattach examples. Keep the existing-subscriber inventory-ready retry alias only at its retained callback, plus retained held-input cancellation/replacement and overlapping attach/loss fences. The inventory driver also loses the whole empty-worktree-publication registration and folder-host/unverifiable scalar examples. Retire the entire dated10-case attach receipt; retain16 independent receipts only in their original scopes. No retained-memory or real-daemon measurement follows from supplied provider inventories. Changed-source prior results do not validate reduced-source bytes. No fresh execution, equivalent replacement, red/green, measured budget, soak, physical provider/PTY, rendered, live remote/mobile or cross-platform result is claimed. Historical prior account (preserved verbatim): Audit1991: Remove initial-snapshot sequence-domain, limited-tail/truncation and unknown-output-sequence aliases whose callbacks are removed. The source retains partial-chunk replay and requested-snapshot retry/admission at supplied snapshot/data seams, not equivalent replacement or measured retained-memory proof. Preserve the existing stream/pending-wait capacity aliases only to their retained owning multiplex callbacks with supplied PTY resolutions. Removing the sibling-PTY heartbeat dedup work-count sample does not establish a new decorative fanout budget; the unrelated-PTY cancellation fence survives. Preserve17 unrelated dated receipts. Changed-source prior results do not validate reduced-source bytes. No fresh execution, equivalent replacement, red/green, measured budget, soak, physical provider/PTY, rendered, live remote/mobile or cross-platform result is claimed. Historical prior account (preserved verbatim): Audit1983: Remove the reasonless-close selector and narrow census aliases to actual retained runtime and injected-inventory callbacks. The removed synthetic census-dedup and isolated pre-boundary timing cases receive no replacement credit; preserve17 unrelated dated receipts. Changed-source prior results do not validate reduced-source bytes. No fresh execution, equivalent replacement, red/green, measured budget, soak, provider/PTY, rendered, live remote/mobile or cross-platform result is claimed. Historical prior account (preserved verbatim): Audit1939 (sources1934/1935): Delete the synthetic watcher-synchronization and partial-reconciliation suites. Retire all affected host-fact, capability-owner, activation-residue and retention-cap aliases because the remaining callbacks do not establish those full claims. Keep exact local hook/timer, byte-stream side effects, exit/disposal, scroll-intent cleanup and deterministic local rank/TTL progression callbacks within their mocked-host scope. No physical paired-host retention, resource budget or equivalent replacement evidence is assigned. The generic broad test command remains a runnable discovery command, never a dated execution receipt. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Delete the complete hook-agent-status projection suite and its two behavioral assertion aliases. Supplied stored hook/title/resume/staleness/rename/question/generation/PTY-retirement/snapshot-version projections are intentionally lost. The separate osc-status-scan gate mentions this source only in existing historical notes and receives no mutation. No measured hidden-output retention, CPU/RSS or resource budget result is claimed by the deleted source or this audit. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Retained exact unchanged-source receipts remain the only current execution evidence. Reduced selectors have no fresh execution result. Historical pre-audit oracle account for changed sources: The retained unchanged-source receipts are the only current execution evidence; reduced-source assertions await validation. Historical oracle account (not a current claim for changed sources): Existing exact unchanged-selector receipts remain the only current execution evidence. Changed source selectors have no fresh execution result. Historical prior-source oracle account: Existing exact unchanged-selector receipts remain the only current execution evidence. Changed source selectors have no fresh execution result. Historical prior-source oracle account: Run one byte-identical six-terminal oracle against an isolated headed desktop host and an isolated headless `orca serve` host. Stage at least 1,000,000 xterm cells, hide all six warm managers, emit an initial working title followed by 30 timed real-PTY spinner frames and a final idle title per worktree, and observe a second real session-tabs subscription plus the production renderer store. Require exactly two transport envelopes and two title mutations per worktree, no decorative title crossing either boundary, and positive decoded-envelope bytes. Enable ordinary parking with the lossy retention budget disabled, require exactly one mounted manager and five parked tabs, at most 45% retained cells, no more than 16 MiB heap growth, and under 500 ms timer drift. In headed mode, require the host renderer to remain on its original workspace with zero target terminal managers mounted throughout client park, reveal, and live I/O. While a tab is parked, require authoritative terminal.read to observe new PTY output; reveal it and require the original PTY, a marker within the requested 5,000-row history, the parked marker, and post-reveal input/output. Drive 24 real PTY title trackers in distinct worktrees through 40 decorative spinner frames each, then 40 spinner frames alternating with Cursor's bare native identity redraw, using fake clocks. Require all 2,880 raw chunks exactly and zero host session-tabs publications, serialized bytes, renderer apply calls, and renderer store mutations. Continue decorative evidence beyond 30 minutes, require exactly one refresh per worktree per 15-minute lease through a FIFO spaced by at least 50 ms, and require all 24 viewer statuses to remain working and fresh; then require one exact semantic idle transition and one visible-output chunk per PTY. Require ordinary and interior-Braille title changes to publish. Drive 64 host PTYs through ten 80 ms synthetic title frames, require zero paired client events after the first frame per PTY while all 640 local frames remain observable, attach a late client and require one current frame per PTY, then require semantic bell and idle transitions on both clients. Reconstruct a legacy subscribed stream with no outputPause capability, hide a chatty pane, require no hidden xterm writes, reveal it, and require one authoritative snapshot plus continued live output. Drop and acknowledge output only for one original paired-client stream, prove the fixture process consumed input while the host model advanced and the client stayed stale, then require the command snapshot probe to replace that stream, repaint exact fixture output, preserve authoritative PTY identity and target-tab cardinality, and resume live I/O. For a client without a delivered sequence, require the first numeric probe to establish a baseline without replacement, a same-sequence probe to remain attached, and a later advanced probe to replace only that stream. Withhold the first encrypted shared-control ready frame, start one RPC, trigger a status-probe refresh, and require two connections, one host delivery, successful response, and zero retained request bytes. Admit 128 active streams, reject the 129th as retryable, release one stream, then require the retry to attach and publish a snapshot without multiplying retained subscribers. Disable terminal.paired-parking.v1 and require the same oracle to fail before parking, while the legacy limit-one fallback separately passes. Also preserve truncated first paint, ACK-starved same-PTY recovery, responsive silent-command snapshot probes, dead-stream probe timeout recovery, and queued manual-disconnect fencing.", + "oracle": "Audit3563 scope: retire the whole Kiro OSC/hook identity suite and the OpenCode2 identity-only case; reduce Codex cached startup child-retirement and missing-session-ID tables to one case each (2 authored/4 finite removals). Intentionally lose Kiro alias/session continuity, OpenCode2 completed OSC identity continuity, cached replacement without relay-generation restart, and missing-previous-ID context-preservation alarms. Remaining provider, remote compatibility, persistence and cross-session pollution cases are not equivalent replacements. Also remove the private Codex roster-size assertion while retaining public child-absence checks. Historical evidence retains its original source scope; no new execution or transferred pass validates these reductions. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. The decorative fanout suite retains only real OSC9999 ingestion through runtime/title heartbeat/viewer past the stale boundary, preserving stateStartedAt and clearing expired prompt metadata. Spinner volume/zero-work inventories, Pi/Grok frame tables, rich-status and foreground-reset examples are retired. Existing receipts for separate paired-terminal sources retain their original scopes. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The retained runtime registration driver loses three complete imported fragments: changed-working-label title facts and suppressed-frame tracked snapshots; refusing SSH GitLab MR-base lookup before local Git/glab calls; and missing folder-workspace path creation/update refusal before spawn/store mutation. These alarms are intentionally lost. Other imported fragments and separately rooted dated receipts are unchanged; they are not equivalent replacement coverage. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3342 removes three supplied activation/inventory fallback and same-handle retry examples; their fallback/retry alarms are intentionally lost without equivalent replacement. The exact mounted-HUB pending old-stream-end versus replacement-snapshot registration and its assertion remain. Unrelated dated macOS receipts retain only their original scope; no fresh runtime, provider, performance, mobile, mixed-version or cross-platform validation is claimed. Audit3326 adopts reconciled source audit3324 and intentionally retires these low-value examples and their regression alarms: restarts only the stream whose renderer delivery credit never settles — An unreturned renderer delivery credit stalls only its stream, leaving healthy sibling/shared transport alive. Fake-clock deadline plus real multiplexer unsubscribe isolates recovery ownership.; probes then restarts a stream when an entered command receives no frames — Entered command produces no output; actual multiplexer must probe then close only its stalled stream after probe timeout. Distinct response liveness from parser credit.; establishes a probe baseline before recovering an unsequenced stream — 142-line manual five-probe baseline/regression sequence derives request IDs from encoded mock traffic and checks repeated no-close/close choreography. Oversized algorithm/state-machine example; no actual renderer/host command or recovery delivery. Retires unsequenced/unchanged/regressed baseline classification alarms.. Retained tests do not replace these alarms. Dated unrelated evidence remains scoped to its original execution; no historical source/import-closure equivalence, fresh execution, real network/provider/SSH, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Audit 3318 reduces remote-runtime-pty-transport-unpublished-host-graph.test.ts, retiring its direct published-empty-graph omission-to-pane-exit alarm. The unchanged bounded unpublished-frame/pushed-recovery and stale-handle park/manual-retry cases are narrower, not equivalent replacements. Both complete three-case Linux receipts dated 2026-11-04 are archived as historical, including source heads; they do not validate this reduced selector. Remaining fourteen active receipts are macOS only. No fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. The direct side-effect-facts selector retains six cases: replay exclusion, live sequence handling, stale disposer ownership, one handoff drain, drain-before-snapshot, and stale replacement-snapshot refusal. Successful current-entry snapshot application is not witnessed by the retained snapshot case. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2090 (sources2088/2089): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2086 (sources2082/2083): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-handles-part-02.spec.ts: recovers a moved SSH pane through the tab it now sits in, not the one its lease froze; reports a plain expired orphan; does not report a pane whose only recent lease was superseded; does not report a pane whose only recent lease had its relay id recycled; picks the eligible orphan over a superseded predecessor that matches first; collects the same tabs the per-tab reader reports, in one sweep of the leases; reports no candidate when no lease names any of the worktree tabs. src/main/runtime/orca-runtime-tests/runtime-availability.spec.ts: starts unavailable with no authoritative window; reports the configured Windows terminal shell on status; reports floating workspace availability from settings on status; polls floating tabs with targeted PTY liveness and no repo/provider inventory; hydrates persisted tabs when the store cannot report repos; advertises browser screencast only when a renderer window is available; routes mobile Codex reset consumption through the account mutation coordinator; maps a definite pre-provider rejection into an authoritative current snapshot. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2003: Remove SSH/absent-inventory attach refusal, later-subscriber retry, no-view adoption/takeback, provider-tail/unknown-read projections, failed-keystroke retry and simpler loss reattach examples. Keep the existing-subscriber inventory-ready retry alias only at its retained callback, plus retained held-input cancellation/replacement and overlapping attach/loss fences. The inventory driver also loses the whole empty-worktree-publication registration and folder-host/unverifiable scalar examples. Retire the entire dated10-case attach receipt; retain16 independent receipts only in their original scopes. No retained-memory or real-daemon measurement follows from supplied provider inventories. Changed-source prior results do not validate reduced-source bytes. No fresh execution, equivalent replacement, red/green, measured budget, soak, physical provider/PTY, rendered, live remote/mobile or cross-platform result is claimed. Historical prior account (preserved verbatim): Audit1991: Remove initial-snapshot sequence-domain, limited-tail/truncation and unknown-output-sequence aliases whose callbacks are removed. The source retains partial-chunk replay and requested-snapshot retry/admission at supplied snapshot/data seams, not equivalent replacement or measured retained-memory proof. Preserve the existing stream/pending-wait capacity aliases only to their retained owning multiplex callbacks with supplied PTY resolutions. Removing the sibling-PTY heartbeat dedup work-count sample does not establish a new decorative fanout budget; the unrelated-PTY cancellation fence survives. Preserve17 unrelated dated receipts. Changed-source prior results do not validate reduced-source bytes. No fresh execution, equivalent replacement, red/green, measured budget, soak, physical provider/PTY, rendered, live remote/mobile or cross-platform result is claimed. Historical prior account (preserved verbatim): Audit1983: Remove the reasonless-close selector and narrow census aliases to actual retained runtime and injected-inventory callbacks. The removed synthetic census-dedup and isolated pre-boundary timing cases receive no replacement credit; preserve17 unrelated dated receipts. Changed-source prior results do not validate reduced-source bytes. No fresh execution, equivalent replacement, red/green, measured budget, soak, provider/PTY, rendered, live remote/mobile or cross-platform result is claimed. Historical prior account (preserved verbatim): Audit1939 (sources1934/1935): Delete the synthetic watcher-synchronization and partial-reconciliation suites. Retire all affected host-fact, capability-owner, activation-residue and retention-cap aliases because the remaining callbacks do not establish those full claims. Keep exact local hook/timer, byte-stream side effects, exit/disposal, scroll-intent cleanup and deterministic local rank/TTL progression callbacks within their mocked-host scope. No physical paired-host retention, resource budget or equivalent replacement evidence is assigned. The generic broad test command remains a runnable discovery command, never a dated execution receipt. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Delete the complete hook-agent-status projection suite and its two behavioral assertion aliases. Supplied stored hook/title/resume/staleness/rename/question/generation/PTY-retirement/snapshot-version projections are intentionally lost. The separate osc-status-scan gate mentions this source only in existing historical notes and receives no mutation. No measured hidden-output retention, CPU/RSS or resource budget result is claimed by the deleted source or this audit. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Retained exact unchanged-source receipts remain the only current execution evidence. Reduced selectors have no fresh execution result. Historical pre-audit oracle account for changed sources: The retained unchanged-source receipts are the only current execution evidence; reduced-source assertions await validation. Historical oracle account (not a current claim for changed sources): Existing exact unchanged-selector receipts remain the only current execution evidence. Changed source selectors have no fresh execution result. Historical prior-source oracle account: Existing exact unchanged-selector receipts remain the only current execution evidence. Changed source selectors have no fresh execution result. Historical prior-source oracle account: Run one byte-identical six-terminal oracle against an isolated headed desktop host and an isolated headless `orca serve` host. Stage at least 1,000,000 xterm cells, hide all six warm managers, emit an initial working title followed by 30 timed real-PTY spinner frames and a final idle title per worktree, and observe a second real session-tabs subscription plus the production renderer store. Require exactly two transport envelopes and two title mutations per worktree, no decorative title crossing either boundary, and positive decoded-envelope bytes. Enable ordinary parking with the lossy retention budget disabled, require exactly one mounted manager and five parked tabs, at most 45% retained cells, no more than 16 MiB heap growth, and under 500 ms timer drift. In headed mode, require the host renderer to remain on its original workspace with zero target terminal managers mounted throughout client park, reveal, and live I/O. While a tab is parked, require authoritative terminal.read to observe new PTY output; reveal it and require the original PTY, a marker within the requested 5,000-row history, the parked marker, and post-reveal input/output. Drive 24 real PTY title trackers in distinct worktrees through 40 decorative spinner frames each, then 40 spinner frames alternating with Cursor's bare native identity redraw, using fake clocks. Require all 2,880 raw chunks exactly and zero host session-tabs publications, serialized bytes, renderer apply calls, and renderer store mutations. Continue decorative evidence beyond 30 minutes, require exactly one refresh per worktree per 15-minute lease through a FIFO spaced by at least 50 ms, and require all 24 viewer statuses to remain working and fresh; then require one exact semantic idle transition and one visible-output chunk per PTY. Require ordinary and interior-Braille title changes to publish. Drive 64 host PTYs through ten 80 ms synthetic title frames, require zero paired client events after the first frame per PTY while all 640 local frames remain observable, attach a late client and require one current frame per PTY, then require semantic bell and idle transitions on both clients. Reconstruct a legacy subscribed stream with no outputPause capability, hide a chatty pane, require no hidden xterm writes, reveal it, and require one authoritative snapshot plus continued live output. Drop and acknowledge output only for one original paired-client stream, prove the fixture process consumed input while the host model advanced and the client stayed stale, then require the command snapshot probe to replace that stream, repaint exact fixture output, preserve authoritative PTY identity and target-tab cardinality, and resume live I/O. For a client without a delivered sequence, require the first numeric probe to establish a baseline without replacement, a same-sequence probe to remain attached, and a later advanced probe to replace only that stream. Withhold the first encrypted shared-control ready frame, start one RPC, trigger a status-probe refresh, and require two connections, one host delivery, successful response, and zero retained request bytes. Admit 128 active streams, reject the 129th as retryable, release one stream, then require the retry to attach and publish a snapshot without multiplying retained subscribers. Disable terminal.paired-parking.v1 and require the same oracle to fail before parking, while the legacy limit-one fallback separately passes. Also preserve truncated first paint, ACK-starved same-PTY recovery, responsive silent-command snapshot probes, dead-stream probe timeout recovery, and queued manual-disconnect fencing.", "commands": [ "pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-unpublished-host-graph.test.ts", "ORCA_BACKGROUND_LAUNCH=1 pnpm exec playwright test tests/e2e/paired-remote-terminal-host-quit-reconnect-input.spec.ts --config tests/playwright.config.ts --project electron-headless --workers=1", @@ -8571,6 +8572,7 @@ "Keep the budget-off control, exact manager count, final bounded-tail flood marker, parked-time output, and post-reveal PTY input in both topology runs." ], "knownGaps": [ + "Audit3563 scope: retire the whole Kiro OSC/hook identity suite and the OpenCode2 identity-only case; reduce Codex cached startup child-retirement and missing-session-ID tables to one case each (2 authored/4 finite removals). Intentionally lose Kiro alias/session continuity, OpenCode2 completed OSC identity continuity, cached replacement without relay-generation restart, and missing-previous-ID context-preservation alarms. Remaining provider, remote compatibility, persistence and cross-session pollution cases are not equivalent replacements. Also remove the private Codex roster-size assertion while retaining public child-absence checks. Historical evidence retains its original source scope; no new execution or transferred pass validates these reductions.", "NEXT65 current scope qualification: The NEXT65 knownGaps entries below govern the changed selectors and qualify any broader coverage implication in this gate’s unchanged oracle/coverageNotes narratives. Those narratives and preserved dated evidenceRuns retain their historical text and original exact execution scopes only; no fresh execution, transferred pass or equivalent replacement coverage is claimed. Upstream case removals remain upstream; original NEXT65 proposal descriptions grant no duplicate retirement credit.", "NEXT65 terminal-owner layout and whole-case policy: src/renderer/src/components/terminal-pane/parking/output-watchers/terminal-parked-watcher-reconciliation.test.ts. Complete retired/reduced selectors lose their listed alarms; retained cases have no transferred historical pass or equivalent replacement coverage. Unaffected dated receipts retain only their original exact scopes. This metadata edit is not fresh execution evidence.", "NEXT65 historical original active selector fields (verbatim JSON values; not current coverage):\ncommands: [\n \"pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-unpublished-host-graph.test.ts\",\n \"ORCA_BACKGROUND_LAUNCH=1 pnpm exec playwright test tests/e2e/paired-remote-terminal-host-quit-reconnect-input.spec.ts --config tests/playwright.config.ts --project electron-headless --workers=1\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/main/runtime/terminal/query/runtime-terminal-inventory-contracts.test.ts\",\n \"ORCA_BACKGROUND_LAUNCH=1 pnpm exec playwright test tests/e2e/paired-client-first-terminal-unopened-host-worktree.spec.ts --config tests/playwright.config.ts --project electron-headless --workers=1\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/main/runtime/session-lifecycle/publication/mobile-session-tabs-agent-status-heartbeat.test.ts tests/e2e/session-tabs-decorative-title-fanout.unit.test.ts --maxWorkers=1\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/components/terminal-pane/parking/lifecycle/use-terminal-tab-cold-parking.test.ts --maxWorkers=1\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/terminal-multiplex-initial-snapshot-buffering.test.ts src/main/runtime/rpc/terminal-multiplex-snapshot-serialization.test.ts src/main/runtime/rpc/terminal-multiplex-pty-wait-capacity.test.ts src/main/runtime/rpc/terminal-subscribe-buffer.test.ts src/renderer/src/components/terminal-pane/pty-transport/output/terminal-side-effect-facts-handler.test.ts src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-hidden-replay.test.ts src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-visibility-and-exit.test.ts src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-activation-inventory-fallback.test.ts src/renderer/src/components/terminal-pane/parking/retention/terminal-hidden-view-parking.test.ts src/renderer/src/components/terminal-pane/parking/retention/terminal-hidden-worktree-retention.test.ts src/renderer/src/components/terminal-pane/parking/output-watchers/terminal-parked-tab-watchers.test.ts src/renderer/src/components/terminal-pane/parking/output-watchers/terminal-parked-watcher-reconciliation.test.ts src/renderer/src/runtime/rpc/subscriptions/runtime-client-events.test.ts src/renderer/src/web/web-preload-api-runtime-environment.test.ts src/main/ipc/runtime-environments/runtime-environments-pairing.test.ts\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/main/runtime/terminal/adoption/terminal-subscriber-driven-daemon-attach.test.ts\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/main/runtime/session-tabs/runtime-inventory-publication-authority.test.ts src/main/runtime/rpc/methods/session-tabs/session-tabs-inventory-census-race.test.ts src/main/daemon/terminal-host/testing/terminal-host-agent-session.test.ts tests/e2e/session-tabs-empty-inventory-daemon-oracle.unit.test.ts --maxWorkers=1\",\n \"pnpm exec vitest run --config config/vitest.config.ts src/shared/remote-runtime/control/remote-runtime-shared-control-connection.test.ts\",\n \"SKIP_BUILD=1 pnpm exec playwright test tests/e2e/paired-remote-terminal-host-restart-background-sync.spec.ts --config tests/playwright.config.ts --project electron-headless --workers=1\",\n \"ORCA_E2E_FORWARD_APP_LOGS=1 SKIP_BUILD=1 pnpm exec playwright test tests/e2e/paired-remote-terminal-host-restart-background-sync.spec.ts --config tests/playwright.config.ts --project electron-headless --workers=1 --repeat-each=10 --max-failures=1\",\n \"ORCA_E2E_WEB_CLIENT=1 SKIP_BUILD=1 pnpm exec playwright test tests/e2e/paired-remote-terminal-truncated-tail-first-paint.spec.ts --config tests/playwright.config.ts --project electron-headful --workers=1\",\n \"SKIP_BUILD=1 pnpm exec playwright test tests/e2e/paired-remote-terminal-lossy-initial-snapshot.spec.ts --config tests/playwright.config.ts --project electron-headful --workers=1\",\n \"ORCA_E2E_WEB_CLIENT=1 ORCA_E2E_DISABLE_PAIRED_TERMINAL_PARKING=1 SKIP_BUILD=1 pnpm exec playwright test tests/e2e/paired-remote-terminal-truncated-tail-first-paint.spec.ts --config tests/playwright.config.ts --project electron-headful --workers=1\",\n \"ORCA_E2E_WEB_CLIENT=1 SKIP_BUILD=1 pnpm exec playwright test tests/e2e/paired-remote-terminal-stall-recovery.spec.ts --config tests/playwright.config.ts --project electron-headful --workers=1\",\n \"ORCA_E2E_WEB_CLIENT=1 SKIP_BUILD=1 pnpm exec playwright test tests/e2e/paired-remote-terminal-probe-gap-recovery.spec.ts --config tests/playwright.config.ts --project electron-headful --workers=1\",\n \"ORCA_E2E_WEB_CLIENT=1 SKIP_BUILD=1 pnpm exec playwright test tests/e2e/headless-paired-remote-terminal-stall-recovery.spec.ts --config tests/playwright.config.ts --project electron-headful --workers=1\",\n \"ORCA_E2E_WEB_CLIENT=1 pnpm exec playwright test tests/e2e/paired-remote-terminal-retention-memory.spec.ts --config tests/playwright.config.ts --project electron-headful --workers=1\",\n \"ORCA_E2E_WEB_CLIENT=1 SKIP_BUILD=1 pnpm exec playwright test tests/e2e/paired-remote-terminal-retention-memory.spec.ts --config tests/playwright.config.ts --project electron-headful --workers=1\",\n \"ORCA_E2E_WEB_CLIENT=1 SKIP_BUILD=1 pnpm exec playwright test tests/e2e/headless-paired-remote-terminal-retention-memory.spec.ts --config tests/playwright.config.ts --project electron-headless --workers=1\",\n \"SKIP_BUILD=1 pnpm exec playwright test tests/e2e/terminal-parked-memory.spec.ts --config tests/playwright.config.ts --project electron-headful --workers=1\"\n ]\ntestFiles: [\n \"src/main/runtime/session-lifecycle/publication/mobile-session-tabs-agent-status-heartbeat.test.ts\",\n \"tests/e2e/session-tabs-decorative-title-fanout.unit.test.ts\",\n \"src/renderer/src/components/terminal-pane/parking/lifecycle/use-terminal-tab-cold-parking.test.ts\",\n \"src/main/runtime/construction/orca-runtime.test.ts\",\n \"src/renderer/src/web/web-preload-api-runtime-environment.test.ts\",\n \"src/main/ipc/runtime-environments/runtime-environments-pairing.test.ts\",\n \"src/main/runtime/rpc/terminal-multiplex-initial-snapshot-buffering.test.ts\",\n \"src/main/runtime/rpc/terminal-multiplex-snapshot-serialization.test.ts\",\n \"src/main/runtime/rpc/terminal-multiplex-pty-wait-capacity.test.ts\",\n \"src/main/runtime/rpc/terminal-subscribe-buffer.test.ts\",\n \"src/main/runtime/terminal/adoption/terminal-subscriber-driven-daemon-attach.test.ts\",\n \"src/main/runtime/session-tabs/runtime-inventory-publication-authority.test.ts\",\n \"src/main/runtime/rpc/methods/session-tabs/session-tabs-inventory-census-race.test.ts\",\n \"src/main/daemon/terminal-host/testing/terminal-host-agent-session.test.ts\",\n \"tests/e2e/session-tabs-empty-inventory-daemon-oracle.unit.test.ts\",\n \"src/renderer/src/components/terminal-pane/pty-transport/output/terminal-side-effect-facts-handler.test.ts\",\n \"src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-hidden-replay.test.ts\",\n \"src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-visibility-and-exit.test.ts\",\n \"src/renderer/src/components/terminal-pane/parking/retention/terminal-hidden-worktree-retention.test.ts\",\n \"src/renderer/src/components/terminal-pane/parking/retention/terminal-hidden-view-parking.test.ts\",\n \"src/renderer/src/components/terminal-pane/parking/output-watchers/terminal-parked-tab-watchers.test.ts\",\n \"src/renderer/src/components/terminal-pane/parking/output-watchers/terminal-parked-watcher-reconciliation.test.ts\",\n \"src/renderer/src/runtime/rpc/subscriptions/runtime-client-events.test.ts\",\n \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-activation-inventory-fallback.test.ts\",\n \"src/shared/remote-runtime/control/remote-runtime-shared-control-connection.test.ts\",\n \"tests/e2e/paired-remote-terminal-lossy-initial-snapshot.spec.ts\",\n \"tests/e2e/paired-remote-terminal-truncated-tail-first-paint.spec.ts\",\n \"tests/e2e/paired-remote-terminal-stall-recovery.spec.ts\",\n \"tests/e2e/paired-remote-terminal-probe-gap-recovery.spec.ts\",\n \"tests/e2e/headless-paired-remote-terminal-stall-recovery.spec.ts\",\n \"tests/e2e/paired-remote-terminal-retention-memory.spec.ts\",\n \"tests/e2e/headless-paired-remote-terminal-retention-memory.spec.ts\",\n \"tests/e2e/terminal-parked-memory.spec.ts\",\n \"tests/e2e/paired-remote-terminal-host-restart-background-sync.spec.ts\",\n \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-unpublished-host-graph.test.ts\",\n \"tests/e2e/paired-remote-terminal-host-quit-reconnect-input.spec.ts\",\n \"src/main/runtime/terminal/query/runtime-terminal-inventory-contracts.test.ts\",\n \"tests/e2e/paired-client-first-terminal-unopened-host-worktree.spec.ts\"\n ]\nassertionRefs: [\n {\n \"file\": \"tests/e2e/session-tabs-decorative-title-fanout.unit.test.ts\",\n \"assertions\": [\n \"2,880 decorative spinner and bare-native real-PTY OSC chunks publish zero session-tabs snapshots or serialized bytes after semantic convergence\",\n \"the viewer performs zero snapshot applies and zero store mutations during decorative activity\",\n \"after 30 minutes, one globally spaced refresh per worktree keeps every mirrored working status fresh\",\n \"exact Pi and rotating normalized Pi/Grok frames renew explicit and fallback status without changing stateStartedAt or acknowledgement semantics\",\n \"matching title evidence renews retained OSC hook status without resetting stateStartedAt, then drops unverifiable prompt metadata after the hook row expires\",\n \"renderer-authored rich status keeps its original lease, then renews only coarse title-confirmed state and provider identity\",\n \"ordinary-shell title transitions replace working stateStartedAt with the completion transition time\",\n \"decorative heartbeats remain blocked until unknown foreground ownership resolves, then resume at the bounded interval\",\n \"tracked terminal state reset clears delayed foreground ownership markers\",\n \"raw chunks, semantic idle transitions, and visible output remain exact\"\n ]\n },\n {\n \"file\": \"src/main/runtime/terminal/adoption/terminal-subscriber-driven-daemon-attach.test.ts\",\n \"assertions\": [\"retries an existing subscriber when provider inventory becomes ready\"]\n },\n {\n \"file\": \"src/main/runtime/rpc/methods/session-tabs/session-tabs-inventory-census-race.test.ts\",\n \"assertions\": [\n \"creation and removal at the census release edge are subsumed once, and a delayed pre-boundary change is not replayed\",\n \"post-boundary creation and removal replay after the initial snapshot in sequence order\",\n \"create/remove/recreate transitions are preserved without collapse\",\n \"a post-boundary projected-state change is delivered without a new snapshot revision\",\n \"an older coalesced callback cannot overwrite a newer immediate change\",\n \"initialization churn publishes only the latest projected state and repeated structural overflow fails after bounded recollection\",\n \"disconnect aborts the census and removes the real runtime listener\"\n ]\n },\n {\n \"file\": \"tests/e2e/session-tabs-empty-inventory-daemon-oracle.unit.test.ts\",\n \"assertions\": [\n \"unpublished empty inventory keeps resume dispatch at zero with the original daemon session, process id, and writer as the sole live writer\",\n \"an authoritative empty inventory settles the mirror and releases exactly one resume dispatch with zero liveness probes\",\n \"re-enabling unconditional empty hydration makes the identical oracle red with one resume dispatch, two session ids, two process ids, and two writer calls\"\n ]\n },\n {\n \"file\": \"tests/e2e/paired-remote-terminal-host-restart-background-sync.spec.ts\",\n \"assertions\": [\n \"preserves daemon PID and target PTY identity across the paired host relaunch\",\n \"foregrounds the daemon-backgrounded target on its first post-relaunch remote observation\",\n \"delivers target and post-relaunch control terminal input and live output\"\n ]\n },\n {\n \"file\": \"tests/e2e/paired-remote-terminal-probe-gap-recovery.spec.ts\",\n \"assertions\": [\"replaces a stale paired stream when the PTY snapshot advanced\"]\n },\n {\n \"file\": \"tests/e2e/paired-remote-terminal-stall-recovery.spec.ts\",\n \"assertions\": [\n \"restarts one ACK-starved paired terminal stream without replacing its PTY\"\n ]\n },\n {\n \"file\": \"tests/e2e/headless-paired-remote-terminal-stall-recovery.spec.ts\",\n \"assertions\": [\"recovers an ACK-starved stream from an isolated headless Orca host\"]\n },\n {\n \"file\": \"tests/e2e/paired-remote-terminal-lossy-initial-snapshot.spec.ts\",\n \"assertions\": [\n \"a nonempty lossy initial snapshot paints exactly once on a paired client\",\n \"the authoritative host cursor advances and live output resumes exactly once\"\n ]\n },\n {\n \"file\": \"tests/e2e/paired-remote-terminal-truncated-tail-first-paint.spec.ts\",\n \"assertions\": [\n \"paints a paired remote terminal when only its retained text tail overflowed\",\n \"legacy paired hosts retain the lossy hidden-manager budget fallback\"\n ]\n },\n {\n \"file\": \"tests/e2e/paired-remote-terminal-retention-memory.spec.ts\",\n \"assertions\": [\n \"ordinary-parks paired terminals and restores authoritative host scrollback\",\n \"warm-mounted hidden paired streams schedule zero renderer output work under sustained host output\",\n \"six hidden real PTYs publish and apply only exact working and idle title transitions across paired transport\",\n \"client recovery completes while every target terminal remains unmounted in the host renderer\"\n ]\n },\n {\n \"file\": \"tests/e2e/headless-paired-remote-terminal-retention-memory.spec.ts\",\n \"assertions\": [\n \"ordinary-parks paired terminals against an isolated headless Orca host\",\n \"cold-activates only visible paired terminals against an isolated headless host\",\n \"the headless host bounds real-PTY decorative title transport and production store application\",\n \"the headless host shares the zero-hidden-renderer-work and exact-restoration contract\"\n ]\n },\n {\n \"file\": \"tests/e2e/terminal-parked-memory.spec.ts\",\n \"assertions\": [\n \"releases un-parkable hidden worktree buffers only once the retention budget engages\"\n ]\n },\n {\n \"file\": \"src/main/runtime/rpc/terminal-multiplex-pty-wait-capacity.test.ts\",\n \"assertions\": [\n \"admits 128 active streams, rejects the 129th, and reuses released capacity\",\n \"reserves PTY wait capacity independently from active streams\"\n ]\n },\n {\n \"file\": \"src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-hidden-replay.test.ts\",\n \"assertions\": [\n \"pauses capable paired output while hidden and restores exactly on reveal\",\n \"locally gates hidden paired output when a legacy host cannot pause it\"\n ]\n },\n {\n \"file\": \"src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-activation-inventory-fallback.test.ts\",\n \"assertions\": [\n \"keeps a mounted HUB mirror alive when the old stream ends before the replacement snapshot\"\n ]\n },\n {\n \"file\": \"src/renderer/src/web/web-preload-api-runtime-environment.test.ts\",\n \"assertions\": [\n \"keeps pairing while manual disconnect fences passive reconnects\",\n \"returns a disconnect envelope when a queued active runtime call disconnects\",\n \"returns a disconnect envelope when a queued selected environment call disconnects\"\n ]\n }\n ]", @@ -8851,7 +8853,7 @@ "providers": ["local", "daemon", "ssh", "remote-runtime"], "coveredPlatforms": ["macos"], "coveredProviders": [], - "coverageNotes": "Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3460 final-v2 composes accepted AUTHOR3454/PEER3455 on actual post263 main: intentionally removes99 authored/106 potential expanded cases across16 selected drivers, including7 whole suites, retaining24 authored/26 potential cases. No equivalent replacement or fresh runtime is claimed. Broad root-unit discovery loses whole cross-client workspace-sync, terminal-preview, picker-skill-hint and close-routing suites plus reduced worker-stop, remote-owned-record, scanner, mobile-bundle and daemon-health samples. Separate E2E discovery also loses the paired mobile/desktop mirror journey and two unit drivers. Retained callback/table/outer-loop/guard bytes are exact; retained checks do not replace the removed alarms. No app, native device, mixed-version, remote or cross-platform execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The retained runtime registration driver loses three complete imported fragments: changed-working-label title facts and suppressed-frame tracked snapshots; refusing SSH GitLab MR-base lookup before local Git/glab calls; and missing folder-workspace path creation/update refusal before spawn/store mutation. These alarms are intentionally lost. Other imported fragments and separately rooted dated receipts are unchanged; they are not equivalent replacement coverage. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Module layout SSH policy: Retire the multiplexer private-clock saturation registration, the writer saturation transition-count registration and the changed-source dated writer/backpressure receipt. Retained byte delivery, lane order, fairness, settlement and cancellation cases are unvalidated by that historical receipt. No equivalent replacement, new execution, measured performance, live SSH/WSL or cross-platform proof is claimed. The unchanged WSL protocol receipt retains only its original command scope. Historical prior account (preserved verbatim): Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2090 (sources2088/2089): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2086 (sources2082/2083): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-handles-part-02.spec.ts: recovers a moved SSH pane through the tab it now sits in, not the one its lease froze; reports a plain expired orphan; does not report a pane whose only recent lease was superseded; does not report a pane whose only recent lease had its relay id recycled; picks the eligible orphan over a superseded predecessor that matches first; collects the same tabs the per-tab reader reports, in one sweep of the leases; reports no candidate when no lease names any of the worktree tabs. src/main/runtime/orca-runtime-tests/runtime-availability.spec.ts: starts unavailable with no authoritative window; reports the configured Windows terminal shell on status; reports floating workspace availability from settings on status; polls floating tabs with targeted PTY liveness and no repo/provider inventory; hydrates persisted tabs when the store cannot report repos; advertises browser screencast only when a renderer window is available; routes mobile Codex reset consumption through the account mutation coordinator; maps a definite pre-provider rejection into an authoritative current snapshot. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2070 (sources2068/2069): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime.test.ts: ordinary registration cleanup removes only await import('./orca-runtime-tests/hooks-and-hosted-review.spec'). src/main/ssh/ssh-multiplexer-transport-writer.test.ts: reports saturated epochs and their drain boundary exactly once. src/main/runtime/orca-runtime-tests/worktree-removal-archive-hook-gate.spec.ts: removes and records the waiver when the failure is explicitly overridden; removes without an override record when the hook succeeds; removes when the hook is configured but not requested; removes when no archive hook is configured; holds a desktop removal in the same repo until this one is accepted; joins a desktop removal of the same worktree accepted while this one waited. src/main/runtime/orca-runtime-tests/worktree-ps-agent-row-dismissal.spec.ts: drops the row as soon as the user dismisses it, without waiting for the PTY to exit; tells paired clients to republish on the transition and on the dismissal; rejoins a row through its %s handle after pane ownership clears. src/main/runtime/orca-runtime-tests/hooks-and-hosted-review.spec.ts: reports an error when the SSH filesystem provider is unavailable; reports an SSH orca.yaml as an error without reading any filesystem; reports ok for a local repo hook check; reads SSH issue commands as absent and drops writes without touching any filesystem; allows host integration slug helpers for SSH repos through provider-aware GitHub clients; routes runtime GitHub repo identity helpers through the selected WSL project runtime; routes runtime GitHub issue and work-item actions through the selected WSL project runtime; pins explicit origin preference on runtime open-by-number work item lookups; forwards check-details cancellation without local Git overrides. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. The historical receipt/currentCommand selecting src/main/ssh/ssh-multiplexer-transport-writer.test.ts remains historical and is not a passing result for the reduced current command. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Delete the complete tail-wait memo suite: its sole current authored callback checks supplied fromTail/blocked field values. Retire all three broader memoization/equivalence/eviction assertion aliases; they are not backed by retained source callbacks. No scan-work budget, sustained output, physical terminal or resource measurement is assigned to this removed test. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1674 removes low-value test projections, mock wiring, private-state assertions, inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained public behavior and refusal boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1634 removes low-value IPC, filesystem and runtime projections, argument inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained cancellation, ownership, authorization, filesystem and SQLite boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Local macOS evidence covers main-process pending-output and projection-admission caps, saturated-tail wait detection, direct-SSH source retention/accounting, always-on V1 negotiation with legacy peer fallback, exact provider-generation pause ownership, renderer exit/data ordering, decoder input bounds, bounded reconnect scheduling, and WSL stdio transport settlement through deterministic unit contracts. A joined main/runtime oracle exercises renderer-sourced headed semantics and headless-model snapshot semantics without launching a live paired server. Separately, a macOS-hosted Docker OpenSSH run exercises only the deployed Linux relay and direct SSH provider with an exact 256 KiB source-credit plateau, concurrent PTY typing, fixed-size filesystem frames, Git churn, and owner-lease reconnect. Neither deterministic topology labels nor Docker SSH constitute live headed desktop, headless orca serve, or physical WSL evidence; local/daemon, prior-version daemons, Windows named pipes/ConPTY, folder workspaces, and mixed-version clients also remain uncovered. Audit1312 removes low-value imported specs from aggregate test drivers; old whole-driver execution receipts are archived. Retained registrations are unchanged. No equivalent replacement coverage is claimed. Audit1383: src/main/runtime/orca-runtime-tests/worktree-native-chat-setup.spec.ts reduced. Lose command-list/receipt/default/background activation matrices whose provider spawn returns canned IDs. Retained test proves actual tab allocation separation, not terminal startup. No equivalent replacement coverage is claimed. Existing exact unaffected receipts and their platform/provider/budget claims remain scoped to unchanged selectors. Changed optional selectors have no new execution result. Audit1401: src/main/runtime/orca-runtime-tests/mobile-session-tabs-duplicate-pty-repair.spec.ts removed. Loses listing-on-already-repaired-data alarm; it never established runtime repair invocation. src/main/runtime/orca-runtime-tests/mobile-session-tabs-split-agent-tab.spec.ts removed. Loses actual mobile split sibling shell versus recorded-agent resume command alarms; no equivalent coverage asserted. src/main/runtime/orca-runtime-tests/worktree-native-chat-setup.spec.ts removed. Loses separate setup/startup tab result in that manually assembled path. src/main/runtime/orca-runtime.test.ts reduced. Driver registration/import of a whole-deleted selected spec is removed; directly authored cases remain unchanged. No extra authored removal credit. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result.", + "coverageNotes": "Audit3563 scope: retire the whole Kiro OSC/hook identity suite and the OpenCode2 identity-only case; reduce Codex cached startup child-retirement and missing-session-ID tables to one case each (2 authored/4 finite removals). Intentionally lose Kiro alias/session continuity, OpenCode2 completed OSC identity continuity, cached replacement without relay-generation restart, and missing-previous-ID context-preservation alarms. Remaining provider, remote compatibility, persistence and cross-session pollution cases are not equivalent replacements. Also remove the private Codex roster-size assertion while retaining public child-absence checks. Historical evidence retains its original source scope; no new execution or transferred pass validates these reductions. Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3460 final-v2 composes accepted AUTHOR3454/PEER3455 on actual post263 main: intentionally removes99 authored/106 potential expanded cases across16 selected drivers, including7 whole suites, retaining24 authored/26 potential cases. No equivalent replacement or fresh runtime is claimed. Broad root-unit discovery loses whole cross-client workspace-sync, terminal-preview, picker-skill-hint and close-routing suites plus reduced worker-stop, remote-owned-record, scanner, mobile-bundle and daemon-health samples. Separate E2E discovery also loses the paired mobile/desktop mirror journey and two unit drivers. Retained callback/table/outer-loop/guard bytes are exact; retained checks do not replace the removed alarms. No app, native device, mixed-version, remote or cross-platform execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The retained runtime registration driver loses three complete imported fragments: changed-working-label title facts and suppressed-frame tracked snapshots; refusing SSH GitLab MR-base lookup before local Git/glab calls; and missing folder-workspace path creation/update refusal before spawn/store mutation. These alarms are intentionally lost. Other imported fragments and separately rooted dated receipts are unchanged; they are not equivalent replacement coverage. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Module layout SSH policy: Retire the multiplexer private-clock saturation registration, the writer saturation transition-count registration and the changed-source dated writer/backpressure receipt. Retained byte delivery, lane order, fairness, settlement and cancellation cases are unvalidated by that historical receipt. No equivalent replacement, new execution, measured performance, live SSH/WSL or cross-platform proof is claimed. The unchanged WSL protocol receipt retains only its original command scope. Historical prior account (preserved verbatim): Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2090 (sources2088/2089): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2086 (sources2082/2083): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-handles-part-02.spec.ts: recovers a moved SSH pane through the tab it now sits in, not the one its lease froze; reports a plain expired orphan; does not report a pane whose only recent lease was superseded; does not report a pane whose only recent lease had its relay id recycled; picks the eligible orphan over a superseded predecessor that matches first; collects the same tabs the per-tab reader reports, in one sweep of the leases; reports no candidate when no lease names any of the worktree tabs. src/main/runtime/orca-runtime-tests/runtime-availability.spec.ts: starts unavailable with no authoritative window; reports the configured Windows terminal shell on status; reports floating workspace availability from settings on status; polls floating tabs with targeted PTY liveness and no repo/provider inventory; hydrates persisted tabs when the store cannot report repos; advertises browser screencast only when a renderer window is available; routes mobile Codex reset consumption through the account mutation coordinator; maps a definite pre-provider rejection into an authoritative current snapshot. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2070 (sources2068/2069): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime.test.ts: ordinary registration cleanup removes only await import('./orca-runtime-tests/hooks-and-hosted-review.spec'). src/main/ssh/ssh-multiplexer-transport-writer.test.ts: reports saturated epochs and their drain boundary exactly once. src/main/runtime/orca-runtime-tests/worktree-removal-archive-hook-gate.spec.ts: removes and records the waiver when the failure is explicitly overridden; removes without an override record when the hook succeeds; removes when the hook is configured but not requested; removes when no archive hook is configured; holds a desktop removal in the same repo until this one is accepted; joins a desktop removal of the same worktree accepted while this one waited. src/main/runtime/orca-runtime-tests/worktree-ps-agent-row-dismissal.spec.ts: drops the row as soon as the user dismisses it, without waiting for the PTY to exit; tells paired clients to republish on the transition and on the dismissal; rejoins a row through its %s handle after pane ownership clears. src/main/runtime/orca-runtime-tests/hooks-and-hosted-review.spec.ts: reports an error when the SSH filesystem provider is unavailable; reports an SSH orca.yaml as an error without reading any filesystem; reports ok for a local repo hook check; reads SSH issue commands as absent and drops writes without touching any filesystem; allows host integration slug helpers for SSH repos through provider-aware GitHub clients; routes runtime GitHub repo identity helpers through the selected WSL project runtime; routes runtime GitHub issue and work-item actions through the selected WSL project runtime; pins explicit origin preference on runtime open-by-number work item lookups; forwards check-details cancellation without local Git overrides. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. The historical receipt/currentCommand selecting src/main/ssh/ssh-multiplexer-transport-writer.test.ts remains historical and is not a passing result for the reduced current command. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Delete the complete tail-wait memo suite: its sole current authored callback checks supplied fromTail/blocked field values. Retire all three broader memoization/equivalence/eviction assertion aliases; they are not backed by retained source callbacks. No scan-work budget, sustained output, physical terminal or resource measurement is assigned to this removed test. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1674 removes low-value test projections, mock wiring, private-state assertions, inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained public behavior and refusal boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1634 removes low-value IPC, filesystem and runtime projections, argument inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained cancellation, ownership, authorization, filesystem and SQLite boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Local macOS evidence covers main-process pending-output and projection-admission caps, saturated-tail wait detection, direct-SSH source retention/accounting, always-on V1 negotiation with legacy peer fallback, exact provider-generation pause ownership, renderer exit/data ordering, decoder input bounds, bounded reconnect scheduling, and WSL stdio transport settlement through deterministic unit contracts. A joined main/runtime oracle exercises renderer-sourced headed semantics and headless-model snapshot semantics without launching a live paired server. Separately, a macOS-hosted Docker OpenSSH run exercises only the deployed Linux relay and direct SSH provider with an exact 256 KiB source-credit plateau, concurrent PTY typing, fixed-size filesystem frames, Git churn, and owner-lease reconnect. Neither deterministic topology labels nor Docker SSH constitute live headed desktop, headless orca serve, or physical WSL evidence; local/daemon, prior-version daemons, Windows named pipes/ConPTY, folder workspaces, and mixed-version clients also remain uncovered. Audit1312 removes low-value imported specs from aggregate test drivers; old whole-driver execution receipts are archived. Retained registrations are unchanged. No equivalent replacement coverage is claimed. Audit1383: src/main/runtime/orca-runtime-tests/worktree-native-chat-setup.spec.ts reduced. Lose command-list/receipt/default/background activation matrices whose provider spawn returns canned IDs. Retained test proves actual tab allocation separation, not terminal startup. No equivalent replacement coverage is claimed. Existing exact unaffected receipts and their platform/provider/budget claims remain scoped to unchanged selectors. Changed optional selectors have no new execution result. Audit1401: src/main/runtime/orca-runtime-tests/mobile-session-tabs-duplicate-pty-repair.spec.ts removed. Loses listing-on-already-repaired-data alarm; it never established runtime repair invocation. src/main/runtime/orca-runtime-tests/mobile-session-tabs-split-agent-tab.spec.ts removed. Loses actual mobile split sibling shell versus recorded-agent resume command alarms; no equivalent coverage asserted. src/main/runtime/orca-runtime-tests/worktree-native-chat-setup.spec.ts removed. Loses separate setup/startup tab result in that manually assembled path. src/main/runtime/orca-runtime.test.ts reduced. Driver registration/import of a whole-deleted selected spec is removed; directly authored cases remain unchanged. No extra authored removal credit. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result.", "motivatingLinks": [ "https://github.com/stablyai/orca/pull/6836", "https://github.com/stablyai/orca/pull/6858", @@ -8859,7 +8861,7 @@ "https://github.com/stablyai/orca/pull/7054" ], "invariant": "High-volume terminal output must stay bounded across daemon socket writes, SSH relay writer/decoder/source retention/reconnect, main model and projection admission, runtime metadata, detectors, and tail checks, main-to-renderer in-flight bytes, renderer scheduler queues, and hidden-output restore without starving focused input. Every new SSH session offers V1 and every same-build relay supports it; legacy delivery is reachable only when capability negotiation proves the peer cannot use V1. Negotiated SSH spans remain exact and contiguous through activation, exit, cancellation, recovery, desktop projection, and required remote replacement; provisional activation data cannot project, reconnect may release it only into the exact attempt's private recovery quarantine after contract validation, and exit seals that quarantine against later same-token frames until ordered admission or exact cancellation proof. Reconnect checkpoint capture freezes exact provider-generation/PTy admission, cancels queued old work, and waits for the running raw completion before exposing its accepted checkpoint; timeout or failure detaches only that PTy's old model, makes the checkpoint unavailable, and releases the bounded fence without closing the shared provider, while the same failure outside an active exact migration remains generation-fatal and overlapping reconnects preserve any earlier outstanding fence. Pending renderer projection IDs are capped and compacted across split remainders, exit preparation owns its renderer fence through finalization, exit-time cancellation transfers published projections before proof commit, generation close fences late proof while draining exact projection waiters, closed-generation identity compacts without weakening stale rejection, canceled source-delivery retirement retains at most one ordered token per PTY, and pause/resume targets only the exact provider generation. Additional subscribers cannot stall the owner, while the required legacy primary retains backpressure.", - "oracle": "Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3460 final-v2 composes accepted AUTHOR3454/PEER3455 on actual post263 main: intentionally removes99 authored/106 potential expanded cases across16 selected drivers, including7 whole suites, retaining24 authored/26 potential cases. No equivalent replacement or fresh runtime is claimed. Broad root-unit discovery loses whole cross-client workspace-sync, terminal-preview, picker-skill-hint and close-routing suites plus reduced worker-stop, remote-owned-record, scanner, mobile-bundle and daemon-health samples. Separate E2E discovery also loses the paired mobile/desktop mirror journey and two unit drivers. Retained callback/table/outer-loop/guard bytes are exact; retained checks do not replace the removed alarms. No app, native device, mixed-version, remote or cross-platform execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The retained runtime registration driver loses three complete imported fragments: changed-working-label title facts and suppressed-frame tracked snapshots; refusing SSH GitLab MR-base lookup before local Git/glab calls; and missing folder-workspace path creation/update refusal before spawn/store mutation. These alarms are intentionally lost. Other imported fragments and separately rooted dated receipts are unchanged; they are not equivalent replacement coverage. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Module layout SSH policy: Retire the multiplexer private-clock saturation registration, the writer saturation transition-count registration and the changed-source dated writer/backpressure receipt. Retained byte delivery, lane order, fairness, settlement and cancellation cases are unvalidated by that historical receipt. No equivalent replacement, new execution, measured performance, live SSH/WSL or cross-platform proof is claimed. The unchanged WSL protocol receipt retains only its original command scope. Historical prior account (preserved verbatim): Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2090 (sources2088/2089): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2086 (sources2082/2083): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-handles-part-02.spec.ts: recovers a moved SSH pane through the tab it now sits in, not the one its lease froze; reports a plain expired orphan; does not report a pane whose only recent lease was superseded; does not report a pane whose only recent lease had its relay id recycled; picks the eligible orphan over a superseded predecessor that matches first; collects the same tabs the per-tab reader reports, in one sweep of the leases; reports no candidate when no lease names any of the worktree tabs. src/main/runtime/orca-runtime-tests/runtime-availability.spec.ts: starts unavailable with no authoritative window; reports the configured Windows terminal shell on status; reports floating workspace availability from settings on status; polls floating tabs with targeted PTY liveness and no repo/provider inventory; hydrates persisted tabs when the store cannot report repos; advertises browser screencast only when a renderer window is available; routes mobile Codex reset consumption through the account mutation coordinator; maps a definite pre-provider rejection into an authoritative current snapshot. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2070 (sources2068/2069): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime.test.ts: ordinary registration cleanup removes only await import('./orca-runtime-tests/hooks-and-hosted-review.spec'). src/main/ssh/ssh-multiplexer-transport-writer.test.ts: reports saturated epochs and their drain boundary exactly once. src/main/runtime/orca-runtime-tests/worktree-removal-archive-hook-gate.spec.ts: removes and records the waiver when the failure is explicitly overridden; removes without an override record when the hook succeeds; removes when the hook is configured but not requested; removes when no archive hook is configured; holds a desktop removal in the same repo until this one is accepted; joins a desktop removal of the same worktree accepted while this one waited. src/main/runtime/orca-runtime-tests/worktree-ps-agent-row-dismissal.spec.ts: drops the row as soon as the user dismisses it, without waiting for the PTY to exit; tells paired clients to republish on the transition and on the dismissal; rejoins a row through its %s handle after pane ownership clears. src/main/runtime/orca-runtime-tests/hooks-and-hosted-review.spec.ts: reports an error when the SSH filesystem provider is unavailable; reports an SSH orca.yaml as an error without reading any filesystem; reports ok for a local repo hook check; reads SSH issue commands as absent and drops writes without touching any filesystem; allows host integration slug helpers for SSH repos through provider-aware GitHub clients; routes runtime GitHub repo identity helpers through the selected WSL project runtime; routes runtime GitHub issue and work-item actions through the selected WSL project runtime; pins explicit origin preference on runtime open-by-number work item lookups; forwards check-details cancellation without local Git overrides. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. The historical receipt/currentCommand selecting src/main/ssh/ssh-multiplexer-transport-writer.test.ts remains historical and is not a passing result for the reduced current command. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Delete the complete tail-wait memo suite: its sole current authored callback checks supplied fromTail/blocked field values. Retire all three broader memoization/equivalence/eviction assertion aliases; they are not backed by retained source callbacks. No scan-work budget, sustained output, physical terminal or resource measurement is assigned to this removed test. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1674 retires changed-source assertion and execution scope without equivalent replacement coverage. Existing unchanged-selector receipts remain scoped to their original sources. Historical prior-source oracle: Existing exact unchanged-selector receipts remain the only current execution evidence. Changed source selectors have no fresh execution result. Historical prior-source oracle account: The current executable slice injects main-process renderer backlog pressure, then asserts ordinary terminal-wait detection without a joined-tail allocation and with full prompt-family continuity, per-PTY and total pending-output caps, a 1,024-ID projection cap with split-remainder compaction, preserved sequenced-tail metadata, active-pending protection ahead of background trimming, exact provider-generation pause ownership, renderer exit/data ordering, and ACK-gated in-flight bounds. The negotiation seam proves every initial connection and reconnect offers V1, the same-build relay advertises V1 without launch flags, old clients remain token-free, and method-not-found peers fall back without installing source ACK publication. The joined source-intake/multiplex oracle blocks renderer or headless serialization, admits one snapshot-covered span and one trailing span, requires replacement reservation only after the authoritative sequence is known, advances upstream ACK eligibility exactly to the covered boundary after SnapshotEnd, and delivers then ACKs the trailing span through the ordinary live path without cancellation. Source-range ledger and multiplex seams additionally require cumulative byte ACK 40 then 100 for one 100-byte frame to release exact byte credit without early source settlement, admit a contiguous higher-generation recovered token while the prior token remains unsettled, and reject stale stream or source generations without releasing in-flight byte credit. Replacement reservation rolls back every earlier span on a later failure, commit requires the exact frozen transfer states, and an authoritatively reclaimed covered span rejects commit without local trim authority while rollback removes the reservation idempotently. The exit-deadline oracle publishes a source-backed desktop projection, advances a fake clock to cancellation, and requires one projection transfer, proof commit, preparation, and final exit with no provider close, retained obligation, duplicate finalization, or process-lifetime cancellation tombstone; a second controlled proof remains pending across generation close and must never publish final exit, while both paths release the renderer preparation lease exactly once. The model-migration oracle accepts span A, blocks span B's raw emulator completion, begins disconnect migration, and proves no checkpoint or reconnect attach can remain at A; releasing B advances the exact checkpoint to B and projects B once. Its fake-clock arm reaches the 10-second deadline, requires checkpoint-unavailable, one exact model reset, zero retained admission charge and timers, and no effect from late raw settlement. Its two-PTY failure arm rejects one running raw callback under an active exact migration and requires one checkpoint-unavailable result, one model reset, zero provider closes, retained sibling checkpoint and transport liveness, zero charges and timers, and an unaffected other generation; a non-migrating sibling oracle still requires generation close. The intake additionally closes 2,048 sequential provider generations into one exact range, rejects stale events across that range, and preserves an out-of-order live gap until its own close. Direct-SSH contracts additionally prove activation-response settlement before the first source frame, claim-gated projection under same-decoder-turn response/data delivery, exact early-ACK reservation through send settlement or same-token retry, recovery completion visible before lease-held frames transfers those frames only into the private quarantine and admits no output until exact checkpoint-to-recoveryEnd coverage, invalid-checkpoint restore retention through response settlement followed by fresh-token retry and one live source frame, isolated saturated-subscriber eviction with healthy owner liveness, preserved legacy-primary backpressure, one-write ownership across write(false)/drain, control selection ahead of queued ordinary PTY input with FIFO lanes and bounded fairness, liveness rebasing, failed-exit retention through late ACK or exact recovery, response-settlement cancellation authority, token-scoped exit cancellation, exact checkpoint-to-recoveryEnd continuity, stale-owner one-shot retry, no physical PTY teardown on recovery failure, exact provider-generation closure and once-only cleanup when recovery cancellation publication or proof rejects, 16 MiB-plus-1 MiB decoder caps, charged per-PTY/session retention budgets, and eight-wide isolated reattach. The live Electron perf oracle adds hidden-output floods, renderer scheduler queue depth, dropped-output-zero normal scenarios, and active key latency budgets before promotion.", + "oracle": "Audit3563 scope: retire the whole Kiro OSC/hook identity suite and the OpenCode2 identity-only case; reduce Codex cached startup child-retirement and missing-session-ID tables to one case each (2 authored/4 finite removals). Intentionally lose Kiro alias/session continuity, OpenCode2 completed OSC identity continuity, cached replacement without relay-generation restart, and missing-previous-ID context-preservation alarms. Remaining provider, remote compatibility, persistence and cross-session pollution cases are not equivalent replacements. Also remove the private Codex roster-size assertion while retaining public child-absence checks. Historical evidence retains its original source scope; no new execution or transferred pass validates these reductions. Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3460 final-v2 composes accepted AUTHOR3454/PEER3455 on actual post263 main: intentionally removes99 authored/106 potential expanded cases across16 selected drivers, including7 whole suites, retaining24 authored/26 potential cases. No equivalent replacement or fresh runtime is claimed. Broad root-unit discovery loses whole cross-client workspace-sync, terminal-preview, picker-skill-hint and close-routing suites plus reduced worker-stop, remote-owned-record, scanner, mobile-bundle and daemon-health samples. Separate E2E discovery also loses the paired mobile/desktop mirror journey and two unit drivers. Retained callback/table/outer-loop/guard bytes are exact; retained checks do not replace the removed alarms. No app, native device, mixed-version, remote or cross-platform execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The retained runtime registration driver loses three complete imported fragments: changed-working-label title facts and suppressed-frame tracked snapshots; refusing SSH GitLab MR-base lookup before local Git/glab calls; and missing folder-workspace path creation/update refusal before spawn/store mutation. These alarms are intentionally lost. Other imported fragments and separately rooted dated receipts are unchanged; they are not equivalent replacement coverage. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Module layout SSH policy: Retire the multiplexer private-clock saturation registration, the writer saturation transition-count registration and the changed-source dated writer/backpressure receipt. Retained byte delivery, lane order, fairness, settlement and cancellation cases are unvalidated by that historical receipt. No equivalent replacement, new execution, measured performance, live SSH/WSL or cross-platform proof is claimed. The unchanged WSL protocol receipt retains only its original command scope. Historical prior account (preserved verbatim): Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2090 (sources2088/2089): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2086 (sources2082/2083): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-handles-part-02.spec.ts: recovers a moved SSH pane through the tab it now sits in, not the one its lease froze; reports a plain expired orphan; does not report a pane whose only recent lease was superseded; does not report a pane whose only recent lease had its relay id recycled; picks the eligible orphan over a superseded predecessor that matches first; collects the same tabs the per-tab reader reports, in one sweep of the leases; reports no candidate when no lease names any of the worktree tabs. src/main/runtime/orca-runtime-tests/runtime-availability.spec.ts: starts unavailable with no authoritative window; reports the configured Windows terminal shell on status; reports floating workspace availability from settings on status; polls floating tabs with targeted PTY liveness and no repo/provider inventory; hydrates persisted tabs when the store cannot report repos; advertises browser screencast only when a renderer window is available; routes mobile Codex reset consumption through the account mutation coordinator; maps a definite pre-provider rejection into an authoritative current snapshot. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2070 (sources2068/2069): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime.test.ts: ordinary registration cleanup removes only await import('./orca-runtime-tests/hooks-and-hosted-review.spec'). src/main/ssh/ssh-multiplexer-transport-writer.test.ts: reports saturated epochs and their drain boundary exactly once. src/main/runtime/orca-runtime-tests/worktree-removal-archive-hook-gate.spec.ts: removes and records the waiver when the failure is explicitly overridden; removes without an override record when the hook succeeds; removes when the hook is configured but not requested; removes when no archive hook is configured; holds a desktop removal in the same repo until this one is accepted; joins a desktop removal of the same worktree accepted while this one waited. src/main/runtime/orca-runtime-tests/worktree-ps-agent-row-dismissal.spec.ts: drops the row as soon as the user dismisses it, without waiting for the PTY to exit; tells paired clients to republish on the transition and on the dismissal; rejoins a row through its %s handle after pane ownership clears. src/main/runtime/orca-runtime-tests/hooks-and-hosted-review.spec.ts: reports an error when the SSH filesystem provider is unavailable; reports an SSH orca.yaml as an error without reading any filesystem; reports ok for a local repo hook check; reads SSH issue commands as absent and drops writes without touching any filesystem; allows host integration slug helpers for SSH repos through provider-aware GitHub clients; routes runtime GitHub repo identity helpers through the selected WSL project runtime; routes runtime GitHub issue and work-item actions through the selected WSL project runtime; pins explicit origin preference on runtime open-by-number work item lookups; forwards check-details cancellation without local Git overrides. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. The historical receipt/currentCommand selecting src/main/ssh/ssh-multiplexer-transport-writer.test.ts remains historical and is not a passing result for the reduced current command. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Delete the complete tail-wait memo suite: its sole current authored callback checks supplied fromTail/blocked field values. Retire all three broader memoization/equivalence/eviction assertion aliases; they are not backed by retained source callbacks. No scan-work budget, sustained output, physical terminal or resource measurement is assigned to this removed test. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1674 retires changed-source assertion and execution scope without equivalent replacement coverage. Existing unchanged-selector receipts remain scoped to their original sources. Historical prior-source oracle: Existing exact unchanged-selector receipts remain the only current execution evidence. Changed source selectors have no fresh execution result. Historical prior-source oracle account: The current executable slice injects main-process renderer backlog pressure, then asserts ordinary terminal-wait detection without a joined-tail allocation and with full prompt-family continuity, per-PTY and total pending-output caps, a 1,024-ID projection cap with split-remainder compaction, preserved sequenced-tail metadata, active-pending protection ahead of background trimming, exact provider-generation pause ownership, renderer exit/data ordering, and ACK-gated in-flight bounds. The negotiation seam proves every initial connection and reconnect offers V1, the same-build relay advertises V1 without launch flags, old clients remain token-free, and method-not-found peers fall back without installing source ACK publication. The joined source-intake/multiplex oracle blocks renderer or headless serialization, admits one snapshot-covered span and one trailing span, requires replacement reservation only after the authoritative sequence is known, advances upstream ACK eligibility exactly to the covered boundary after SnapshotEnd, and delivers then ACKs the trailing span through the ordinary live path without cancellation. Source-range ledger and multiplex seams additionally require cumulative byte ACK 40 then 100 for one 100-byte frame to release exact byte credit without early source settlement, admit a contiguous higher-generation recovered token while the prior token remains unsettled, and reject stale stream or source generations without releasing in-flight byte credit. Replacement reservation rolls back every earlier span on a later failure, commit requires the exact frozen transfer states, and an authoritatively reclaimed covered span rejects commit without local trim authority while rollback removes the reservation idempotently. The exit-deadline oracle publishes a source-backed desktop projection, advances a fake clock to cancellation, and requires one projection transfer, proof commit, preparation, and final exit with no provider close, retained obligation, duplicate finalization, or process-lifetime cancellation tombstone; a second controlled proof remains pending across generation close and must never publish final exit, while both paths release the renderer preparation lease exactly once. The model-migration oracle accepts span A, blocks span B's raw emulator completion, begins disconnect migration, and proves no checkpoint or reconnect attach can remain at A; releasing B advances the exact checkpoint to B and projects B once. Its fake-clock arm reaches the 10-second deadline, requires checkpoint-unavailable, one exact model reset, zero retained admission charge and timers, and no effect from late raw settlement. Its two-PTY failure arm rejects one running raw callback under an active exact migration and requires one checkpoint-unavailable result, one model reset, zero provider closes, retained sibling checkpoint and transport liveness, zero charges and timers, and an unaffected other generation; a non-migrating sibling oracle still requires generation close. The intake additionally closes 2,048 sequential provider generations into one exact range, rejects stale events across that range, and preserves an out-of-order live gap until its own close. Direct-SSH contracts additionally prove activation-response settlement before the first source frame, claim-gated projection under same-decoder-turn response/data delivery, exact early-ACK reservation through send settlement or same-token retry, recovery completion visible before lease-held frames transfers those frames only into the private quarantine and admits no output until exact checkpoint-to-recoveryEnd coverage, invalid-checkpoint restore retention through response settlement followed by fresh-token retry and one live source frame, isolated saturated-subscriber eviction with healthy owner liveness, preserved legacy-primary backpressure, one-write ownership across write(false)/drain, control selection ahead of queued ordinary PTY input with FIFO lanes and bounded fairness, liveness rebasing, failed-exit retention through late ACK or exact recovery, response-settlement cancellation authority, token-scoped exit cancellation, exact checkpoint-to-recoveryEnd continuity, stale-owner one-shot retry, no physical PTY teardown on recovery failure, exact provider-generation closure and once-only cleanup when recovery cancellation publication or proof rejects, 16 MiB-plus-1 MiB decoder caps, charged per-PTY/session retention budgets, and eight-wide isolated reattach. The live Electron perf oracle adds hidden-output floods, renderer scheduler queue depth, dropped-output-zero normal scenarios, and active key latency budgets before promotion.", "commands": [ "pnpm exec vitest run --config config/vitest.config.ts", "pnpm exec vitest run --config config/vitest.config.ts src/main/runtime/construction/orca-runtime.test.ts", @@ -8958,6 +8960,7 @@ "Keep stress variants non-blocking until stable runtime history exists." ], "knownGaps": [ + "Audit3563 scope: retire the whole Kiro OSC/hook identity suite and the OpenCode2 identity-only case; reduce Codex cached startup child-retirement and missing-session-ID tables to one case each (2 authored/4 finite removals). Intentionally lose Kiro alias/session continuity, OpenCode2 completed OSC identity continuity, cached replacement without relay-generation restart, and missing-previous-ID context-preservation alarms. Remaining provider, remote compatibility, persistence and cross-session pollution cases are not equivalent replacements. Also remove the private Codex roster-size assertion while retaining public child-absence checks. Historical evidence retains its original source scope; no new execution or transferred pass validates these reductions.", "Launch direct-owner cleanup current scope: This gate’s unfiltered unit command now selects a suite reduced by 82 whole registrations / 159 potential table examples in 10 whole test Programs, including 8 registrations / 15 examples from five once-registered create-message factory closures. Lost alarms include frozen first-message/options and legacy-host fallback; distinct actions and redelivery; uncertain/refused/reloaded input custody, Retry and late proof; rich prompt/attachment upload Stop/close and handback; ordinary draft/editor/IME/StrictMode lifetime and accepted topology; provisional fallback tab placement/group/pin/order and rich worktree handoff. Four whole consent/resume-identity cases remain in two Programs; they do not replace these alarms. This entry governs current reduced-unit scope and qualifies any broader implication of unchanged oracle/coverageNotes history. Entire dated evidenceRuns, original commands/currentCommand values, counts/results and existing history retain their original scopes only; no fresh execution, transferred pass, replacement equivalence or prior upstream-removal credit is claimed.", "Audit2145 accepts removal of canned runtime metadata/guard/forwarding, fake in-memory persistence, private work-count and wrapper-duplication tests from the construction/orca-runtime.test.ts aggregate, plus the two automation run-context guards from runtime-workspace-launch-contracts.test.ts. Retained race, framing, physical completion and ownership callbacks remain byte-identical. Existing commands remain valid selectors; historical receipts stay unchanged and provide no fresh validation or replacement coverage for the reduced suites.", "Audit2133 current-parent preparation (988 authored removals; independently reviewed2099/2107/2115/2116 decisions; upstream clear-buffer and recovery-image regressions retained): Intentionally retire these low-value alarms without replacement: src/main/runtime/construction/orca-runtime.test.ts: remove only literal registration import/call pairs await import('../orca-runtime-tests/worktree-setup-and-startup.spec'), await import('../orca-runtime-tests/gitlab-and-pr-bases.spec'); no additional authored credit. src/main/ssh/ssh-channel-multiplexer.test.ts: sends a JSON-RPC request and resolves on response; assigns unique request IDs; sends notifications without expecting a response; dispatches incoming notifications to handler; sends one keepalive frame per cadence tick; drives keepalive sends AND dead-link detection from a single interval; resolves false immediately on a disposed mux; throws on request after dispose; reports isDisposed correctly; clears registered handlers on dispose; does not retain handlers registered after dispose. src/main/persistence/worktree-sort-order-write.test.ts: re-ranks every writable row when a blocked row sits in the requested order. src/main/runtime/orca-runtime-tests/targeting-and-resilience.spec.ts: routes tab switch and capture start by explicit page id; accepts focus on tab switch without altering bridge args (focus is main-side concern); keeps terminal cwd resolution fail-soft when the provider is unavailable. src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-04.spec.ts: drops the bounded scan cache when orphan-cleanup removal completes; clears optimistic reconcile tokens when a CLI worktree removal succeeds; emits drift without mutating when the fetched base fast-forwards created HEAD; emits base_changed without mutation when the fetched base rewrote history; skips stale-token reconciles without mutating or emitting stale status; emits unknown without mutation when fetch fails or the base ref is missing. src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-03.spec.ts: routes runtime preserved-branch force-delete through the selected WSL project runtime; replies to a waiting client with the error of a delete that fails after acceptance; treats forced runtime deletion of an already-missing unregistered worktree as cleanup; treats normal runtime deletion of an already-missing unregistered worktree as cleanup. src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-02.spec.ts: warns that a missing-repo removal only forgot the workspace; restores runtime watchers when CLI worktree deletion fails after teardown; does not re-spend the drain budget on a removal after a wedged install was abandoned; routes runtime worktree removal through the selected WSL project runtime; force-deletes a branch that was preserved by runtime worktree removal. src/main/runtime/orca-runtime-tests/gitlab-and-pr-bases-part-02.spec.ts: records GitLab pasted-project recents only after successful runtime lookup; threads explicit origin preference through runtime WSL PR base resolution; resolves local GitLab fork MR bases from the target project MR head ref; captures the fork MR head from a dedicated ref, not the shared FETCH_HEAD; keeps the durable MR head when the head fetch fails but the local ref resolves; routes runtime GitLab fork MR base git calls through the selected WSL project runtime. src/main/runtime/orca-runtime-tests/gitlab-and-pr-bases.spec.ts: passes SSH connection ids through GitLab task operations; routes runtime GitLab issue, MR, work-item, and todo actions through the selected WSL project runtime; routes runtime GitLab MR details, review-management, job, and pasted URL actions through the selected WSL project runtime; normalizes runtime GitLab issue list arguments like the desktop IPC path. src/main/runtime/orca-runtime-tests/worktree-setup-and-startup-part-03.spec.ts: does not warn when setup is explicitly skipped for CLI-created worktrees; puts the agent a create starts in the first default tab, as the window does; keeps the 8s main-runtime startup readiness budget for agents without an override; gives OpenCode the 20s main-runtime startup readiness budget; launches explicit startup agents with prompts for CLI-created worktrees. src/main/runtime/orca-runtime-tests/worktree-setup-and-startup.spec.ts: returns a setup launch payload for CLI-created worktrees when hooks are explicitly enabled; does not hand the window a setup main never tried to start; passes the selected Windows setup shell into runtime runner generation; follows normal setup policy for CLI-created worktrees without activating them; uses returned WSL setup shell metadata when runtime spawns setup; uses the shell-aware setup runner for windowless creates without a startup command; reports the in-process setup hook as running when nothing can launch the runner. src/main/runtime/orca-runtime-tests/terminal-sleep-and-teardown-part-02.spec.ts: coalesces two clients sleeping the same host worktree; stops exactly the expected live PTYs for a worktree; explains a refused workspace close instead of throwing the raw refusal enum. src/main/runtime/orca-runtime-tests/mobile-summaries-part-03.spec.ts: marks the desktop-active worktree as isActive; includes SSH-backed worktrees in the mobile worktree summary; caches repeated malformed path misses before normalizing them again. src/main/runtime/orca-runtime-tests/mobile-summaries-part-02.spec.ts: keeps title-only foreground work ahead of monitoring in another pane; keeps a local hook row while a connected PTY still backs its pane; keeps an OSC row via its connected PTY after the pane binding is cleared. src/main/runtime/orca-runtime-tests/mobile-summaries.spec.ts: resolves files through a source-visible linked-checkout worktree; marks saved session tabs with live PTYs as host sidebar activity; attributes live legacy PTYs from saved layout bindings when their panes are hidden; marks saved browser tabs as host sidebar activity like desktop; falls back to the path-keyed GitHub cache entry; includes folder workspaces in compact worktree summaries for mobile; attaches inline agent rows from the latest OSC 9999 status; projects monitoring over a title-derived working status. src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-03.spec.ts: pushes to an idle pane when the only live waiter filters out the message type; resolves a registered waiter without PTY-injecting when the leaf is already idle; keeps already-idle status after tui-idle wait for immediate message delivery; resolves message waiters when notifyMessageArrived is called; does not resolve type-filtered message waiters for unrelated message types; removes message waiter abort listeners after message arrival; resolves message waiters on timeout when no message arrives. src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-02.spec.ts: delivers pending mail via notifyMessageArrived when the recipient is already idle; does not keep retrying when every pending row was already pointed; repoints pending rows restored with a live-idle coordinator. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-13.spec.ts: rolls back a half-created terminal whose surface never publishes. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-12.spec.ts: rejects startup prompts for agents that require post-ready stdin; uses portable Unix quoting for mobile agent launch commands in WSL project runtimes; keeps PowerShell quoting for mobile agent launch commands in Windows host runtimes; uses cmd.exe quoting for mobile agent launch commands in local Windows host runtimes; publishes headless mobile session agent identity with synthesized PTY status; rejects disabled mobile session agent launches before spawning; validates mobile terminal insertion anchors before resolving agent launch commands. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-11.spec.ts: coalesces a reconnect close burst onto one authoritative PTY inventory; keeps an explicit user close destructive while the PTY is live; builds mobile session agent launch commands on the runtime host; injects mobile quick-command prompts into the host-built agent startup command. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-10.spec.ts: closes persisted headless terminal parents before any prior list call; tears down a serve-owned headless tab on close while a renderer is attached so it cannot resurrect; defers a renderer-published pending tab to the renderer instead of tearing it down; returns a delegated close outcome with no selection tombstone when the renderer owns the outcome; tears down a runtime pending shell the renderer never adopted on close. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-08.spec.ts: materializes a plain shell when the pending tab has no launch agent; hydrates legacy persisted terminal tabs without layout entries; reads and indexes the full agent-status snapshot once per mobile projection; kills persisted SSH PTYs when closing hydrated headless tabs before pane metadata is restored; materializes hydrated pending headless terminals with the persisted session identity. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-07.spec.ts: hydrates a persisted SSH-owned pane before an attached renderer publishes its graph. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-05.spec.ts: publishes the desktop's tab name over the live agent title, and the live title without one; uses fresh neutral PTY titles over stale mobile snapshot and OSC titles; operates PTY-backed mobile session terminals without a renderer graph; reuses pane close for live PTYs that do not own a renderer tab; lists PTY-backed mobile session terminals without a renderer graph; shows and resolves active PTY-backed mobile session terminals without a renderer graph. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-04.spec.ts: creates mobile session terminals for folder workspaces in a headless runtime server; spawns fresh headless SSH mobile session terminals instead of reattaching synthetic local ids; hydrates headless mobile session terminals from the host workspace session; keeps live headless mobile session terminals when a desktop renderer publishes without them; publishes laptop-created remote runtime split terminals to phone session tabs. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-02.spec.ts: does not report headless wake degradation without sleeping records; does not wake slept agents for non-mobile session-only activation; does not rewrite unread metadata when a mobile activation finds the worktree already read; returns unread:false from worktree.ps after a mobile activation clears the flag; materializes pending mobile session terminals without focusing desktop clients; closes browser mobile session tabs when addressed by browser workspace id. src/main/runtime/orca-runtime-tests/mobile-session-tabs.spec.ts: resolves mobile terminal surfaces by exact split leaf; closes the matching mobile terminal UUID leaf without closing the whole tab; closes the whole mobile terminal tab when addressed by parent tab id; activates the active split leaf when addressed by parent tab id; clears unread metadata on mobile worktree activation without focusing desktop clients; wakes slept agents on the host renderer when a phone activates a worktree. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-05.spec.ts: publishes the hook provider session on a headless mobile tab so native chat can address the transcript; recovers the agent type from the hook row when the pane was launched without an agent hint; reads one agent-status snapshot per projection, not one per terminal tab; waits for unknown-launch foreground owner before publishing Pi-compatible mobile status; keeps Pi-compatible mobile session status as Pi for an unknown-launch foreground pi PTY; keeps renderer-vetted mobile agent status for custom-titled terminals. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-04.spec.ts: publishes terminal surface agent status for paired web clients; preserves authoritative OMP identity for Pi-compatible remote terminal snapshots; normalizes a remote OMP title without republishing omitted launch identity; skips the foreground-process probe when the PTY launch agent is already known; probes the foreground process only on a status transition for unknown launch agents; passes the client signal and the slept-worktree refusal to the spawn. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-03.spec.ts: does not use stale runtime-created PTY status when a neutral PTY title exists; recognizes runtime-created Codex PTY handles from the ready prompt; recognizes runtime-created Antigravity PTY handles from the ready prompt; recognizes live leaf Antigravity terminals from the ready prompt; recognizes a later Antigravity header with a prompt but no model line; keeps mobile terminal surfaces visible while their leaf handle is pending. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-02.spec.ts: lets Claude agents management titles clear stale runtime-created title status; does not recognize live Claude agents panes from a Claude foreground process; does not let stale tab-level agent titles override current neutral pane titles; lists live terminals with fresh pane titles over stale tab titles; lets adopted neutral pane titles use non-shell foreground fallback; lets adopted neutral pane titles retry wrapper foregrounds until recognized; authorizes settled CLI prompts only after positive %s foreground identity; does not recognize arbitrary foreground TUIs as running agents; lets live neutral pane titles retry wrapper foregrounds until recognized; lets current Claude agents PTY titles override stale runtime-created OSC titles; does not let stale Claude agents OSC titles suppress current PTY title activity. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status.spec.ts: keeps preallocated terminal handles valid across renderer reloads; preserves runtime-created PTY process identity after graph unavailable; keeps prompt bindings fenced across runtime restarts without provider incarnation; recognizes runtime-created PTY handles with agent launch titles; does not treat a bare Cursor Agent native title as a running agent session; uses stale runtime-created PTY status when there is no title or foreground evidence. src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-07.spec.ts: keeps orphaned list and show writability aligned with the send gate. src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-04.spec.ts: cancels a coalesced SSH worker recovery retry when its provider disconnects. src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-03.spec.ts: retries renderer reveal before clearing an adopted legacy worker sleeping record. src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-02.spec.ts: recovers exported ORCA_TERMINAL_HANDLE from discovered live PTY sessions. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-12.spec.ts: applies ANSI erase-from-start line controls in retained previews; applies ANSI stripping for private or intermediate CSI line controls; applies ANSI stripping for unsupported erase-line modes; reveals an agent terminal under the pane the caller reserved; keeps a fresh %s pane host-owned when a desktop window is available; keeps desktop publication %s separate from the reserved host spawn; mints its own pane when the reserved leaf is malformed, and reports the one it used; starts an agent terminal with the model the caller picked; starts an agent terminal without a model flag when no pick was made; attributes a startup-agent launch named %s as %s; attributes a fresh agent session the host builds as unknown; leaves a bare agent command the user typed out of launch attribution. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-11.spec.ts: trims oversized terminal output bursts without per-line array shifts; builds terminal previews without mapping the full retained tail; keeps recent PTY replay output capped without needing previous data for large chunks; applies terminal redraw controls before retaining previews; applies ANSI terminal redraw controls before retaining previews; retains multi-line ANSI footer redraws instead of appending old frames; keeps the cursor column when erasing full lines in multi-line redraws; applies ANSI split redraw controls without leaking raw params. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-10.spec.ts: returns renderer visible screen lines through terminal.read RPC JSON result; separates composer drafts from provider-owned terminal screens. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-09.spec.ts: reports an in-flight PTY focus as not navigated when its notifier disappears; reports graph-leaf focus as not navigated without a host notifier; grounds the provider mode tracker before a later chunk, in the emulator order; grounds an in-flight provider snapshot capture so it cannot publish the pre-reset screen; waits for terminal exit and resolves with the exit status; keeps terminal read payloads bounded while retained output remains pageable; trims terminal read preview character budget without per-line array shifts. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-08.spec.ts: submits a silent Claude composer once after the bounded render fallback; chunks large terminal.send text before provider writes; yields chunked terminal input through immediates between writes; yields while validating accepted large terminal.send text before provider writes; reveals a background terminal session when focusing its handle; replays captured launch config when focusing a background agent session; reveals background terminal sessions with the freshest PTY title; renames background terminal handles without requiring a visible tab; keeps a background terminal handle stable while reveal adoption is racing. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-07.spec.ts: resolves tui-idle when a stale Codex prompt is followed by Antigravity readiness; returns an agent-neutral blocked wait result for update prompts; returns an agent-neutral blocked wait result for workspace trust prompts; returns an agent-neutral blocked wait result for cwd selection prompts; returns a blocked wait result for Codex model migration prompts; returns a blocked wait result for Codex startup hook review prompts; returns an agent-neutral blocked wait result for generic interactive prompts; splits text and enter writes for background terminal handles; types the lead line for %s: %s; submits through the agent-specific PTY timing policy for %s. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-06.spec.ts: waits for exit on background terminal handles; keeps retained PTY transcript memory when controller refresh fails; resolves tui-idle from an Antigravity ready prompt preview; resolves Antigravity ready prompts with newline-heavy pasted tails without splitting; resolves live-leaf tui-idle from an Antigravity ready prompt preview. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-03.spec.ts: spawns the available Qoder command for a captured same-session resume: %s; does not use the local Windows shell setting for remote Windows bare agent creates; matches canonical bare agent commands when a command override is configured; keeps non-bare agent command terminal creates unchanged; keeps disabled bare agent command terminal creates unchanged; injects runtime hook receiver env into terminal sessions; creates background terminal sessions for a $label; creates background terminal sessions for a $label; enables Claude Agent Teams only for direct Claude launches when configured in-process; reveals Claude Agent Teams launches with the rewritten launch config; preserves Claude Agent Teams for sequenced Claude launches. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness.spec.ts: creates visible terminal sessions without asking the renderer to focus a tab; asks the pty controller for the requested shell instead of a startup command; knows the agent a flagged launch command names before any process read; retires a flagged command pane identity when the command exits; retires only receipted restored PTY authority on command completion and exit; forgets synthetic handles when disconnected PTY records are pruned. src/main/runtime/orca-runtime-tests/headless-snapshots.spec.ts: returns OSC titles from headless main terminal snapshots; adopts renderer-seeded titles into headless main terminal snapshots; returns cwd metadata from seeded headless main terminal snapshots; adopts OSC7 host metadata from seeded headless terminal scrollback; falls back to the renderer snapshot for hidden-output recovery without headless state; keeps an empty headless snapshot authoritative for hidden-output recovery; advances the absolute output sequence across a daemon stream gap; emits explicit OSC 9999 agent status from runtime PTY data. src/main/runtime/orca-runtime-tests/terminal-side-effect-facts-part-02.spec.ts: falls back to the provider snapshot for a restored PTY with no mounted renderer; falls back to the available mirror when authoritative provider history is unavailable; keeps an empty renderer snapshot when the provider has no retained content; keeps restored provider history authoritative after later live output; uses provider alternate-screen state while a partial model is unsafe; prefers the tracked title over the headless emulator lastTitle; keeps the cursor-agent literal in record-fallback snapshots only without a tracker title; touches mobile snapshots once for decorative spinner ticks, again on idle. src/main/runtime/orca-runtime-tests/local-worktree-creation-part-02.spec.ts: creates a same-repo PR branch override from a resolved head SHA and matching push target; skips broad remote fetch for an existing full-SHA PR base; creates a selected Bitbucket PR branch override from a matching remote branch; suffixes an existing PR when a matching push target lacks selected PR metadata; suffixes a matching push target branch when selected PR metadata has no PR number; suffixes a matching push target branch when the existing PR is different; suffixes a selected PR remote conflict when the PR lookup fails; checks out an unused runtime PR branch only when it is at the resolved head SHA; suffixes only the runtime worktree path when an exact PR branch checkout path exists. src/main/runtime/orca-runtime-tests/local-worktree-creation.spec.ts: returns runtime local base update suggestions from addWorktree; creates a runtime local worktree from a usable persisted local branch base; does not create a runtime local worktree when the refresh fails and no local base ref exists; creates a branchNameOverride worktree from the selected matching remote base ref. src/main/runtime/integrations/skill/orca-runtime-skill-recovery.test.ts: continues skill management after startup recovery fails. src/main/runtime/orchestration/mailbox/routing/orchestration-mailbox-detached-routing.spec.ts: uses bounded ownership lookup for direct arrivals while the agent cannot take input; wakes a bound Run waiter when same-Run direct mail arrives; keeps an unowned detached-recipient message in its direct mailbox; uses primary-key lookups for bounded routing updates. src/main/runtime/orchestration/mailbox/terminal-pointer/orchestration-mailbox-delivery.test.ts: remove only literal registration import/call pairs import './orchestration-mailbox-pointer-cli-command.spec'; no additional authored credit. src/main/runtime/orchestration/mailbox/terminal-pointer/orchestration-mailbox-pointer-cli-command.spec.ts: renders the %s CLI command in a mailbox pointer. src/main/runtime/rpc/methods/terminal/terminal-multiplex-link-estimate.test.ts: sizes a slow high-latency link near twice its bandwidth-delay product; never paces a LAN, even when its client parses slowly; never paces a fast link; returns to legacy windows when the path RTT jumps. src/main/runtime/rpc/methods/terminal/terminal-snapshot-scrollback-rows.test.ts: steps down the row ladder when history is over the byte budget; tries the larger first row count, then the usual ladder within the same byte budget. src/main/runtime/rpc/server/pairing/network/windows-default-route-interfaces.test.ts: returns unknown when route inspection fails or emits malformed data; does not inspect routes outside Windows. src/main/runtime/rpc/server/pairing/pairing-offer-owner-cli.test.ts: is refused to a paired %s client; reports a failed widening instead of advertising an unreachable address. src/main/runtime/rpc/terminal-multiplex-end-verdict.test.ts: reports exited only when the owning runtime completes the exit waiter; reports unverifiable when the runtime cannot observe the exit waiter; reports unverifiable when a disconnected PTY resolves with unknown liveness. src/main/runtime/session-graph/headless-runtime-graph.spec.ts: lets a session tab inventory settle on a host no renderer publishes for. src/main/runtime/session-graph/runtime-graph-standing-contracts.test.ts: remove only literal registration import/call pairs import { registerHeadlessRuntimeGraphTests } from './headless-runtime-graph.spec', registerHeadlessRuntimeGraphTests(); no additional authored credit. src/main/runtime/session-tabs/structured-conversation-tab-title.test.ts: bumps the stored snapshot version only when the title changes. src/main/runtime/session-tabs/terminal-close-observed-exit.test.ts: keeps the healthy aggregate verification and delayed physical exit behavior; uses a stamped exit for the incarnation that was actually being closed. src/main/runtime/terminal/identity/terminal-identity-probe.test.ts: answers live for a structured worker without touching the PTY graph; answers live for a PTY handle the runtime still holds; runs the full leaf check for a handle with no live PTY; answers not-live for a stale handle; propagates \"could not look\" rather than reporting it as a dead handle. src/main/runtime/terminal/layout/paired-desktop-server-writes-layout-owner.test.ts: names the tab in main’s layout through terminal.rename; takes a window-style layout.apply close as already stopped, so the terminal runs on. src/main/runtime/terminal/lifecycle/pty-lifecycle-generation-retention.test.ts: retains no lifecycle generation after a spawn/exit cycle. src/main/runtime/terminal/lifecycle/run-terminal-user-open.test.ts: reads a freshly spawned terminal as used once a user opens its tab, not on a probe; keeps the open mark on the process a user open starts. src/main/runtime/terminal/surface/terminal-surface-replies-window.test.ts: from a desktop window that mounts both panes; from a server with no window. src/main/runtime/terminal/surface/workspace-layout-projection.test.ts: carries views for a partition that holds none yet. src/main/runtime/terminal/view/terminal-subscribe-exit-waiter-leak.test.ts: releases a signalled exit-waiter when the signal aborts. src/main/runtime/terminal/wait/orca-runtime-exact-worker-provider-session.test.ts: selects a local hook status for a local pane. src/main/runtime/terminal/wait/prime-agent-screen-readiness-transcripts.test.ts: keeps the footer and the bare caret on screen mid-turn; %s: the screen alone reads ready during %s. src/main/runtime/workspaces/startup/runtime-worktree-agent-startup.test.ts: uses per-launch arguments and preserves launch telemetry; attributes a startup agent whose caller named no surface as unknown; launches %s bare for an absent or blank prompt; attributes a %s draft launch named %s as %s; merges an automation's extras over the host defaults; threads worktree-create extras into the startup build; returns an enabled requested agent without detecting; returns the default when nothing was requested; detects instead of using the default when the requested agent is disabled; starts no agent when the default is blank; detects an enabled agent on this host when nothing usable was named; starts no agent when detection finds none. src/main/runtime/workspaces/startup/runtime-worktree-startup-readiness.test.ts: pastes and submits through composer readiness, never typing raw input; types the follow-up on WSL or SSH-to-Windows once the agent is in front, as before; cleans up the deadline when the composer was already captured; a draft that is not submitted is pasted on the box cursor, as before; takes the box after the grace when the row never comes, whatever repaints; takes the box at the deadline when the grace would outlast the budget; lets a dialog on screen veto a grace; refuses the deadline inside a pending grace while a shell is in front; takes the box at the deadline inside a pending grace with the agent in front; takes the box at the deadline after a refused grace once the agent is in front. src/main/skills/skill-install-transaction.test.ts: installs, previews unchanged bytes, and updates a clean receipt-owned skill; preserves the old version after a %s filesystem failure. src/main/speech/model-manager-download-resume.test.ts: requests the remaining bytes when a clean range response ends before the file total; bounds a server that advances by pathologically tiny segments forever. src/main/ssh/config/ssh-config-parser.test.ts: parses a basic host block; parses multiple host blocks; skips wildcard-only Host entries; skips Host entries with only pattern characters; parses IdentityFile with ~ expansion; parses Windows-style IdentityFile with ~ expansion; parses quoted scalar values like OpenSSH; parses equals-form scalar directives; parses IdentityAgent with ~ expansion; parses IdentitiesOnly; parses GSSAPIAuthentication; keeps the first GSSAPIAuthentication value like OpenSSH; parses ProxyCommand, ProxyUseFdpass, and ProxyJump; ignores comments and blank lines; handles case-insensitive keywords; stops current block on Match directive; returns empty array for empty input; creates one parsed host per concrete alias on a multi-pattern Host line; applies identity agent settings to every concrete alias on a multi-pattern Host line; defaults port to 22 for invalid port values; converts hosts to SshTarget objects; uses host alias as hostname when HostName is missing; skips hosts that are already imported; defaults username to empty string when not specified; carries through identityFile, identityAgent, identitiesOnly, proxyCommand, and jumpHost; carries through gssapiAuthentication; imports duplicate aliases only once and keeps the first concrete host; maps hosts for the picker and flags aliases already in Orca; de-dupes aliases and defaults missing fields; parses hostname, user, port from ssh -G output; collects multiple identity files; parses forwardagent yes; defaults port to 22 when missing; returns empty hostname when missing; returns undefined user when missing; handles empty output; skips lines without spaces; parses proxycommand and filters \"none\"; parses proxyusefdpass yes; parses proxyjump and filters \"none\"; handles ~ expansion in identity file paths; handles Windows-style ~ expansion in identity file paths; parses identityagent with ~ expansion; preserves identityagent none so auth can disable agent fallback; parses identitiesonly yes; parses gssapiauthentication; parses controlmaster options and filters controlpath none. src/main/ssh/config/ssh-known-hosts-source.test.ts: splits the space-separated list ssh -G prints on one line; expands ~ in a reported path; keeps a double-quoted path containing spaces whole; falls back to the default files when ssh -G reported nothing; drops an explicit none without falling back to the defaults; splits both known_hosts lists on one line each; reads StrictHostKeyChecking, including accept-new; reads HostKeyAlias and prefers it for the lookup; reads the non-default port; reports that the lookup came from an alias, so the port is not appended; reports no alias when the dialed host is used; unions the entries of every file; skips a path that is a directory; returns nothing when no file can be read; reports nothing unreadable when every file parses; does not count an empty file as unreadable; prefers HostKeyAlias over the resolved hostname; uses the dialed host, because a resolved hostname can just echo the label; still prefers an explicit HostKeyAlias over the dialed host; falls back to the dialed host when nothing was resolved; is rejoined when the split fragments resolve to nothing; leaves a single unresolvable path alone rather than inventing one; expands the __PROGRAMDATA__ token rather than passing it through as a path; does not rewrite a path that merely BEGINS with the token characters; keeps a drive-letter path with mixed separators intact; splits a spaced Windows home into the fragments the rejoin has to put back. src/main/ssh/config/ssh-known-hosts.test.ts: reads the algorithm from the blob rather than trusting the line; rejects %s; refuses a malformed blob: %s; uses the bare host on the default port; tries the bracketed form first and falls back to bare off-port; matches an exact line; matches case-insensitively; reports a changed key of the same type; reports unknown for %s; matches any host in a comma list; matches the bracketed form on its port; does not apply a bracketed entry to the default port; falls back to a bare line when off-port; never reports a change from the fallback pass; matches a bare literal on the default port; matches a bracketed literal on its port; reports a change for a literal; handles %s; still reports a change for a different key on a revoked host; skips a line carrying an unrecognised marker; does not report a change when only another key type is on file; matches within the same type; accepts CRLF terminators; skips blanks, comments and surrounding whitespace; accepts a hit regardless of which file holds it (%s); never brackets the alias, whatever the port; still brackets an ordinary host on a non-default port; matches a bare alias entry on a non-default port; still accepts the unmodified key field; accepts a hashed entry whose salt is a full SHA1 digest; still accepts the unmutated ssh-keygen -H vector; falls back to the bare host line on a non-default port, as ssh did; refuses a key type it has not seen for a host it knows, as ssh did. src/main/ssh/connection/ssh-connect-config.test.ts: returns undefined when no default keys exist; returns the first existing key file; probes regular and FIDO2 keys in stable default order; keeps a regular default ahead of a malformed FIDO2 default; uses target host/port/username; falls back to resolved config when target fields are empty; keeps imported endpoint fields as the fallback when ssh -G is unavailable; sets readyTimeout to CONNECT_TIMEOUT_MS; sets keepaliveInterval to 15s; enables keyboard-interactive auth so MFA challenges can be answered; uses agent auth when no explicit key and SSH_AUTH_SOCK is set; enables agent forwarding when OpenSSH config requests it and an agent is available; does not enable agent forwarding without a usable agent; resolves IdentityAgent SSH_AUTH_SOCK from the environment; uses the Windows OpenSSH agent pipe when no environment socket is available on Windows; wraps agent auth with IdentityFile filtering when IdentitiesOnly is enabled; applies a manual target IdentitiesOnly setting when ssh -G reports it off; includes unencrypted target.identityFile auth when an agent is available; defers encrypted target.identityFile auth when an agent is available; uses keyFile auth when target.identityFile is set and no agent is available; includes unencrypted resolved identityFile auth when an agent is available; defers a default-path resolved identityFile to the no-agent retry; offers the keyboard-interactive challenge when the agent defers no key; provides fallback key when no agent is available; can force private key inclusion for the post-agent fallback path; falls back to resolved proxyCommand; returns structured jump-host config for target.jumpHost; returns structured jump-host config for resolved proxyJump; returns undefined when no proxy is configured. src/main/ssh/connection/ssh-connection-auth-fallback.test.ts: resolves OpenSSH config using configHost when present; tries ssh-agent before reading an explicit private key; falls back to direct private key auth when agent auth fails; falls back to direct private key auth after too many agent authentication failures; answers bounded keyboard-interactive challenges such as Duo 2FA. src/main/ssh/connection/ssh-connection-store.test.ts: lists picker suppression aliases without consulting re-adoption tombstones; getTarget delegates to store; addTarget generates an id and persists; saves the CLI control switch as given, with no consent stamp; addTarget stamps source as manual by default; addTarget preserves an explicitly provided source; inserts a new config host stamped as ssh-config; asks the parser for all hosts — reconciliation happens in the store; updates an existing config-sourced target when the port changed; refreshes gssapiAuthentication on sync; never overwrites a manual target that owns the alias; does not rewrite an unchanged config-sourced target; returns empty array when nothing changed; tombstones a deleted config-sourced host so re-import skips it; suppresses a deleted manual target from config discovery; re-adding a deleted host reclaims its alias so sync stops suppressing it; editing a target reclaims its alias from the deleted set; records a tombstone when removing any user-facing target; re-adopts orphaned repos when the same host is re-added. src/main/ssh/connection/ssh-connection.test.ts: transitions to connected on successful connect; captures the negotiated SSH server key fingerprint; allows concurrent exec commands for ssh2 transport; removes startup listeners after ssh2 connect succeeds; scopes lifecycle events and pending handshake timers to one mock client; enables TCP_NODELAY on the new ssh2 client after a reconnect cycle; transitions through connecting → connected states; reports error state on connection failure; disconnect cleans up and sets state to disconnected; getTarget returns a copy of the target; getState returns a copy of the state; throws when connecting a disposed connection. src/main/ssh/orcad-migration-stuck-host-profiles.test.ts: never leaves a running terminal behind; is offered only when leaving saved state behind would clear the refusal. src/main/ssh/orcad-runtime-maintenance.test.ts: keeps a connect from retrying a build an explicit Update also failed; offers Force when Update was refused by policy; forwards anew after an activation, since the new slot may bind another port; passes a refusal through and leaves the server linked; forwards anew once a serving slot is restored, which may have bound another port. src/main/ssh/port-forwarding/ssh-port-forward.test.ts: reports the port the OS bound when an ssh2 forward asks for port 0; adds an ssh2 port forward and returns entry; adds a system SSH port forward when no ssh2 client is available; reports the port a system SSH forward resolved when asked for port 0; passes resolved OpenSSH config to system SSH port forwards; disposes a system SSH tunnel when removing the forward; lists forwards filtered by connectionId; removes a forward by id; awaits forward close when removing a forward for user actions; returns null when removing nonexistent forward; removes all forwards for a connection; dispose removes all forwards; stores label in the entry; preserves the existing id when updating a forward succeeds. src/main/ssh/system-transport/system-ssh-forward-process.test.ts: spawns exactly the argv the orphan matcher is given, LogLevel=INFO included; runs port forwards unmultiplexed even when ssh_config disables muxing; preserves manual target port and identity options in the forwarded ssh command; does not spawn ssh when the requested local forward port is already in use; spawns ssh after the requested local forward port preflight succeeds; resolves startup after the grace period when ssh keeps running; sends SIGTERM then SIGKILL when the process does not exit; does not resolve stop until the process exits. src/main/ssh/system-transport/system-ssh-loopback-route.test.ts: dials the held forward ahead of the alias so OpenSSH keeps these over ssh_config; drops a manual target's own port and proxies, which would fight the route; never routes a forward, whose argv must stay stable for orphan recovery; %s → %s; opens the forward over the real proxy, then proves a login over the route; lands the forward on the sshd and known_hosts name the direct route resolved; holds the forward with server-alive probes; never lets the background forward or probe ask for credentials; starts no forward when admission refuses the route; stops the forward when sshd refuses logins over it; stops the whole forward process tree on Windows, once; reports a forward that already exited before anyone listened for it. src/main/startup/browser-process-user-agent-ordering.test.ts: explains admission refusal before a desktop launch exits; explains an orcad holding the profile instead of exiting silently; does not acquire profile admission for a duplicate launch; executes after the dev app name and before later preflight work; exits and releases admission after pending profile move recovery fails; keeps Linux background startup failures console-only and immediate; keeps Linux serve startup failures console-only and immediate. src/main/startup/quit/quit-path-durable-write-blocking.test.ts: store.flushAsync() issues no synchronous fs syscalls; store.flushAsync() also moves the active-view and github-cache sidecars off the thread; stats.flushAsync() issues no synchronous fs syscalls; Grok session cleanup issues no synchronous fs syscalls; stats.flushAsync() closes out live agents before it returns, not when its write lands; a stalled mount leaves the event loop live instead of parking it; store.flushAsync() is one idempotent final barrier; store.flushAsync() resolves instead of throwing when the profile mount rejects writes. src/main/stats/agent-session-transition-recorder.test.ts: counts a hook-only agent that never writes an agent-shaped terminal title; never opens a session from a replayed or disk-restored working status; still closes a live session when the terminating status arrives as a replay; counts one session per turn across repeated working/done cycles; closes an open session when its pane is torn down; bounds the mirror and closes what it evicts; stops the session when the main agent settles and only a watch loop holds the row working; keeps the session open while a subagent outlives the main agent, and closes it when the child settles; dates the stop by the evidence when a shell outlives the last subagent; opens a new session dated by the observation when the main agent resumes after monitoring; never dates a reopen before the close a fact-less OSC repaint caused; dates the first live row after a restored one by the evidence, not the restored clock; treats an unchanged answer as a snapshot, never a transition; opens only on a live working edge; closes only a session it opened. src/main/text-generation/commit-message-text-generation-model-discovery.test.ts: returns static catalog models without spawning for static agents; discovers dynamic models through the agent CLI; discovers dynamic models through the configured agent command override; falls back to static models when dynamic discovery returns no parseable models; falls back to the first discovered non-Pi model when its static default is unavailable. src/main/text-generation/pull-request-context.test.ts: does not fetch unrelated fork remotes before generating PR context; bounds exact remote-ref probe concurrency for a large remote list; reports no branch change because PR context preparation is read-only; keeps a remote-qualified base when the selected base includes the remote. src/main/text-generation/source-control-local-process.test.ts: asks the supervisor to stop and leaves its group to it; signals nothing once the supervisor has exited; stops an agent that ignores its stdin end through its supervisor. src/main/updater/linux-package/linux-package-update-recovery.test.ts: retains the downloaded package with the digest from the event metadata; requires a string downloaded path; accepts an uppercase extension; requires a non-empty string version; matches an absolute file URL by basename; accepts duplicate entries that agree on the digest; refuses an entry with a missing or non-string digest; refuses a digest that is not base64 at all; refuses a malformed percent-encoded URL; refuses a missing or non-array file list; tolerates a non-object event; replaces the retained artifact with a newer download; keeps the retained artifact when a new download has no usable digest; keeps the retained artifact when an unrelated download event arrives; clears the retained artifact; preserves the artifact across a recheck of the same version; clears the artifact when another version takes over; preserves the artifact for an unknown or empty version; is a no-op without a retained artifact; returns the built command and package file name for a verified package; reports missing without a retained artifact; propagates a package-manager discovery failure; rejects a well-formed digest that belongs to another file; never hashes a package that a malformed digest failed to arm; rejects a path outside the cache root; rejects the cache root itself; rejects a package in the updater directory itself; rejects a pending directory that is a symlink out of the cache; performs one hash pass for concurrent requests; revalidates once the in-flight pass settles; does not reuse an in-flight pass for a different artifact; starts a fresh proof when the same package is captured again; returns the verified package path; returns missing without a retained artifact. src/main/updater/releases/updater-prerelease-feed-readiness.test.ts: accepts a GitHub %s asset redirect as ready; reports a GitHub asset request error as unavailable; reports not-ready with a verified last-good tag when the newest assets are unavailable; does not classify a non-positive tag limit as feed unavailability; reproduces the v1.4.142 publishing incident as not-ready; reports transport failures as unavailable instead of not-ready; treats an explicit asset 404 as not-ready when another asset is unavailable; accepts absolute manifest asset URLs without rewriting them to release asset paths; returns not-ready with an older ready update as last-good while newest is publishing. src/main/updater/releases/updater-release-builds.test.ts: lists hourly builds from the dedicated repo, newest first; lists daily builds from the dedicated repo, newest first; separates stable from rc in the shared main repo; skips drafts and unparseable tags; keeps a composed release title and drops one that repeats the tag; hides builds that published no artifact for this platform; returns an empty list when no build has this platform artifact; keeps mac builds visible on macOS regardless of the Windows leg; resolves the platform installer download url; leaves the installer url null when the release published no installer; tolerates a release whose assets are missing or malformed; sends the local gh token as a bearer header when one is available; sends no authorization header when gh has no token; retries unauthenticated once when GitHub rejects the token; does not retry a 401 that was already unauthenticated; falls back to the per-IP bucket when the token is rate limited and tells the gh breaker; reports a secondary limit carrying retry-after without retrying or tripping the breaker; records a primary limit in the token scope %s; keeps %s authenticated when only %s is blocked; skips the token while its scope %s is blocked; surfaces a rate limit with its reset time and a sign-in hint when unauthenticated; omits the sign-in hint when the rate-limited request was authenticated; treats 429 as a rate limit; reports a 403 without rate-limit headers as a plain HTTP error; reports a missing hourly repo distinctly; is null when GitHub sent no reset; prefers retry-after over the primary reset epoch; falls back to the primary reset epoch when there is no retry-after; reads retry-after as seconds; reads retry-after as an HTTP date; pins an hourly tag at the hourly repo download path; pins a daily tag at the daily repo download path; pins a stable tag at the main repo download path; rejects a tag that is not a version. src/main/updater/updater.build-channel-selection.test.ts: uses the display label in the unsupported-platform %s pinned-build error; still pins %s from an unsigned Windows dev build; restores ordinary release checks after local build selection fails; opts into the RC channel when checkForUpdatesFromMenu is called with includePrerelease; pins the generic feed to a perf-tagged prerelease when requested; surfaces no-update feedback when no newer perf-tagged prerelease exists; keeps prerelease publishing-window misses on the generic retry path; leaves the feed URL alone for a normal user-initiated check. src/main/updater/updater.check-preflight.test.ts: ignores a stale update-not-available event after a new check starts preflight; deduplicates repeated manual checks while the immediate checking status is active. src/main/updater/updater.prerelease-fallback.test.ts: retries a prerelease check once against the previous feed tag when the manifest is missing; surfaces a promise-only prerelease fallback failure after the primary error event. src/main/updater/updater.publishing-window-feed.test.ts: repins the generic feed to the newest RC tag for a prerelease user; repins the generic feed to a newer stable tag for a prerelease user; pins a stable fallback when Atom has no newer prerelease; uses last-good concrete feed when a user-initiated check lands during publishing; can offer an older fully published update while the newest release is publishing; keeps background publishing-window fallback on the short retry cadence. src/main/updater/updater.quit-and-install.test.ts: surfaces an accepted retry before electron-updater emits download progress; defers quitAndInstall and runs its launcher within the launch PATH scope; runs pre-quit cleanup before local PTY cleanup during update install; keeps optional pre-quit cleanup fail-and-continue behavior by default; ignores duplicate quitAndInstall requests while the shared delay is pending; arms the forced-exit watchdog once the install commits; disarms the forced-exit watchdog when sync install error recovery keeps the app open. src/main/window/createMainWindow-markdown-editor-focus.test.ts: ignores traffic light sync IPC on non-macOS; still intercepts Cmd+Shift+B and Cmd+Alt+B when the markdown editor is focused; coerces non-boolean setMarkdownEditorFocused payloads to false; opens a table-aware context menu synchronously without a renderer query; builds the plain edit menu for the one right-click the renderer marks as plain text. src/main/window/createMainWindow-renderer-crash-recovery.test.ts: notifies the caller when the renderer process is gone; passes the renderer webContents id through crash recording and recovery callbacks; forwards expected renderer teardowns so the recorder can diagnose suppression; reloads the app shell after an unexpected renderer process loss; does not reload after renderer loss when recovery is disabled; stops auto-reloading after a rapid renderer crash loop trips the breaker. src/main/window/dashboard-popout-window.test.ts: updates the title after language initialization and unsubscribes when closed; lets its preload expose saved WSL hosts in serve-only, and passes nothing otherwise; creates a native-frame window with the shared preload and no webview surface; shows the window on ready-to-show; loads the prod file entry; loads the dev server URL when in dev; focuses the existing window instead of creating a second one; trusts only the live popout webContents; restores a minimized window when re-requested; restores valid persisted bounds; discards off-screen persisted bounds and falls back to defaults; persists bounds on resize after the debounce, guarding near-minimum sizes; keeps saving bounds after a vetoed quit and freezes them only on an allowed quit; closeDashboardPopout closes an open window; applies the persisted app zoom level on dom-ready; follows app zoom changes while open and unsubscribes on close; zoomDashboardPopoutIfFocused steps the popout zoom only while focused; handles the zoom-in chord via before-input-event and ignores other keys; handles mouse-wheel zoom requests outside before-input-event; respects zoom keybinding overrides for keyboard and mouse-wheel paths. src/main/window/dropped-path-preparation.test.ts: opens and stores each unique source once as drop size grows tenfold; passes ordinary paths without copying for %s; preserves ordinary successes and copied successes when another file fails; serializes copy requests from different owners; reports all lost paths without a shared reason when their failures differ. src/main/window/main-window-updater-deferred-setup.test.ts: releases all 64 completed fallback handles without repeating initialization; keeps setup after ready-to-show and releases its fallback afterward; preserves the 15-second fallback when first paint never arrives; preserves destroyed-window guard reads before and at the fallback deadline; keeps the exact %s error and releases its completed fallback; keeps the exact fallback %s error without retrying initialization; preserves pending manual setup after failed ready registration; keeps setup options, live store reads, nudge writes and required quit flush. src/main/worker-threads/worker-thread-request-queue.test.ts: posts one request at a time and in the order it was dispatched; fires a posted call at its own deadline; stops respawning and fails the rest of the queue once deaths hit the cap; clears the death count when a fresh burst starts from full idle. src/main/workspace-hooks/hooks-setup-runner-script.test.ts: builds Windows runners for newline-heavy scripts without line-array splitting; builds POSIX runners without regex-wide CRLF normalization; writes POSIX setup runners for shebang-declared scripts on native Windows paths; keeps batch setup scripts on cmd.exe when the terminal is Git Bash; hands a Git Bash pane a launch command MSYS cannot rewrite for a cmd runner; replays interpreter flags declared on the shebang line; preserves cmd.exe setup runner semantics for configured cmd users; keeps POSIX runner behavior on POSIX platforms; waits when either local or project policy requires completed setup; marks setup-runner terminals for the always-on credential guard; maps git-bash to POSIX setup launch metadata on Windows; falls back to the cmd runner when the git-bash setting has no installed Git Bash; keeps the cmd runner for a non-Git bash such as Cygwin; keeps the cmd runner for a bare bash whose flavor cannot be resolved; uses the POSIX runner for a bare bash that resolves to Git Bash; uses the POSIX runner for an extension-less Git Bash path; preserves the existing cmd runner for PowerShell terminals; preserves cmd setup compatibility when a Windows-host project has a WSL shell setting; classifies an installed explicit Git Bash executable as a POSIX runner; falls back to the cmd runner when the explicit Git Bash path no longer exists. src/main/workspaces/catalog/repo-git-remote-identity-enrichment.test.ts: schedules remote identity enrichment without blocking the caller; probes a drive project whose runtime is WSL %s; sweeps a local repo: %s; probes an SSH row that carries only executionHostId on its own host; keeps same-path rows on two different SSH hosts from sharing one backoff; coalesces concurrent probes for the same repo location; caches no-identity probes briefly so list calls do not retry every time; does not rewrite the no-remote marker on a later retry; resolves a settled no-remote repo once it gains a remote; does not re-probe a resolved identity before the refresh window elapses; re-probes a resolved identity after the refresh window elapses; bounds re-probes per sweep so a large repo list drains gradually; forgets a removed repo location so its deadline cannot outlive the repo; keeps a surviving repo backoff when a sibling repo is removed; probes again when the same location is re-added after its probe was retired; coalesces concurrent sweeps instead of running one pass per list call. src/main/workspaces/creation/preparation/worktree-create-preparation.test.ts: preserves create priority through claim probes on %s; starts the checkout only once the async workspace root resolves; still deduplicates when both callers await the same pending root lookup; namespaces native Windows preparation directories for long paths; deduplicates preparation for the same repo, base, runtime, and workspace root; does not spend a divergence walk when the base matches exactly; claims when the two sides spell the same ref differently; evicts a repo's own stale preparation before another repo's; still discards armed checkouts when the project runtime needs repair; routes preparation and finalization through the selected WSL runtime; retries stale cleanup after a transient listing failure; does not discard a detached worktree with caller-controlled preparation metadata; does not re-arm after an isolated create; re-arms a preparation once creates arrive in a burst; holds the re-arm checkout until the create runs the deferred thunk; starts one explicit prefetch after the create completes; does not re-arm when finalization failed; reports a pending create while a stale-cleanup scan is running. src/main/workspaces/removal/local-worktree-removal-recovery.test.ts: recovers Git for Windows partial filesystem deletion failures; finishes a clean non-force deletion after Git leaves a Windows directory behind; recovers localized Windows failures after Git already removed the registration; does not recover unrelated localized failures while Git still owns the row; finishes a macOS deletion Git deregistered before failing with Directory not empty; leaves an orphaned checkout to the orphan cleanup instead of deleting it; prunes and strictly verifies on the selected WSL host without deleting files or branches. src/main/workspaces/removal/worktree-removal-safety.test.ts: rejects deleting a worktree that contains another registered worktree; rejects POSIX home directories even when host homedir has a different path shape; refuses a home the host reported even when no path rule recognises it; refuses the Windows host's home directory from a POSIX client; refuses the Windows host's home directory even when the host reported nothing. src/main/wsl/discovery/distro-catalog.test.ts: retries asynchronous discovery after a transient wsl.exe failure; retries synchronous discovery after a transient wsl.exe failure; retries asynchronous discovery after wsl.exe reports no distros yet; retries synchronous discovery after wsl.exe reports no distros yet; bounds the wsl.exe spawn rate while no distros are installed; bounds the asynchronous spawn rate while no distros are installed; still caches a non-empty distro list for the process lifetime; re-probes a docker-desktop-only machine once a user distro can appear; backs off while the list keeps coming back empty; backs off after repeated distro-list failures; caps the empty-list backoff at five minutes; keeps reporting an empty result after it goes stale; keeps reporting a known distro list indefinitely; bounds cached homes while retaining the most recently used distros; still lists distros, and an excluded distro home probe fails like any other probe; gives the async home probe time for a cold boot instead of killing it at 5 s; retries availability after a probe timeout instead of latching false; bounds the probe rate while the failure window is open; caches a successful probe for the process lifetime; names an explicit spawn directory, so no deleted cwd can read as ENOENT; shares one wsl.exe spawn between concurrent probes; keeps reporting the last observed answer after it goes stale; holds a definitive failure far longer than a timeout when %s; backs off after repeated probe failures; re-probes a seeded retryable failure once its window lapses; re-probes availability once a distro list succeeds after %s; does not drop an availability failure for an empty distro list; keeps reporting unavailable off Windows without probing. src/main/wsl/discovery/wsl-distro-holder.test.ts: runs a busybox-safe loop through --exec and names the distro; spawns a detached-style holder, records it, and waits for the distro to run; a holder exiting with code 1 notes its distro may have stopped (E9); rejects a pid whose command line is not our holder for that distro; cleans up stale and corrupt records; keeps a holder and its record when release cannot verify it; reports unverifiable when the command line is unreadable even with a matching pid; release kills a verified holder and removes its record; release never kills a pid that is no longer our holder; reports a spawned holder exit so liveness can re-check. src/main/wsl/filesystem/wsl-guest-drive-mountinfo.test.ts: skips 9p shares that are not drvfs, and malformed lines. src/main/wsl/orcad/wsl-orcad-deploy-app-update.test.ts: plans the update on the census the running orcad reports, not an unknown one; forwards force to the update planner; does not serve the incumbent past a refusal that is not a wait; does not re-activate the build this host rolled back from; labels the launch with the build the adoption found serving, not the stale record; defers while another run holds the host fence; lets an explicit force deploy past both skips; names a newer host refusal apart from a rolled-back one, so only Force is offered; updates an older host as before, without asking to adopt it; never asks the incumbent when it already runs this build. src/renderer/src/components/agent/agent-launch-picker-acquisition.test.ts: keeps a completed listing usable when the host rejects the wait parameter. src/renderer/src/components/feature-tips/feature-tip-telemetry.test.ts: delivers to chat-default users in all targeting while holding out not-switched users. src/renderer/src/components/native-chat/composer/pickers/use-native-chat-side-chat-command.test.ts: completes the row on a pick, even when the command is already typed; ignores a pick of the row while side chats are unavailable. src/renderer/src/components/native-chat/media/images/native-chat-image-copy.test.ts: explains a too-large refusal that arrives through the IPC error prefix. src/renderer/src/components/native-chat/messages/actions/native-chat-fork-launch.test.ts: keeps the question with the caller and says once that it never went; tells the user to check the side chat when delivery is unconfirmed; reports a delivered question as sent, silently. src/renderer/src/components/native-chat/messages/actions/native-chat-side-chat-fork-point.test.ts: decides each of %i tool-only turns once per layout, and nothing without a listener. src/renderer/src/components/native-chat/messages/actions/use-native-chat-fork.test.tsx: places a side chat whose question was %s: %o. src/renderer/src/components/terminal-pane/provider-notices/agent-presence/agent-terminal-presence.test.ts: calls back once for %s; stops watching once unsubscribed. src/renderer/src/components/terminal-pane/pty-connection/emulation/pty-connection-mode2031-raw-subscription.spec.ts: records a mode 2031 subscribe from the raw chunk boundary without answering. src/renderer/src/components/terminal-pane/pty-connection/input/pty-connection-deferred-input-forwarding.spec.ts: routes terminal input through deferPtyInput so the host can withhold it; forwards terminal input directly when the host supplies no deferPtyInput. src/renderer/src/components/terminal-pane/pty-connection/notifications/pty-connection-terminal-bell-attention.spec.ts: forwards onBell to pane attention and notification policy; hands the bell to the attention module with its pane even when pane attention is off. src/renderer/src/components/terminal-pane/pty-connection/output-restore/pty-connection-dropped-output-snapshot-healing.spec.ts: repaints from the main-owned snapshot when main drops pending output at the cap. src/renderer/src/components/terminal-pane/pty-connection/output/pty-connection-fresh-native-follow-reset.spec.ts: retries a fresh-spawn native follow reset when renderer dimensions return. src/renderer/src/components/terminal-pane/pty-connection/spawn/unbound-pane-spawn-recovery.test.ts: requests a tab remount for an unbound spawn; leaves recovery to a remote transport that armed its own retry; requests a remount when the transport reports no recovery of its own; settles the spawn as failed before requesting a remount; prefers the restored leaf id when reporting the anomaly; falls back to the pane leaf id when no restored leaf exists. src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-foreground-agent.test.ts: remove only literal registration import/call pairs import { registerPtyConnectionDeferredInputForwardingTests } from '../input/pty-connection-deferred-input-forwarding.spec', registerPtyConnectionDeferredInputForwardingTests(); no additional authored credit. src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-hidden-replay.test.ts: structural support changed only; no new authored removal credit. src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-input-and-repaint.test.ts: remove only literal registration import/call pairs import { registerPtyConnectionFreshNativeFollowResetTests } from '../output/pty-connection-fresh-native-follow-reset.spec', registerPtyConnectionFreshNativeFollowResetTests(); no additional authored credit. src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-spawn-and-startup.test.ts: remove only literal registration import/call pairs import { registerPtyConnectionTerminalBellAttentionTests } from '../notifications/pty-connection-terminal-bell-attention.spec', registerPtyConnectionTerminalBellAttentionTests(); no additional authored credit. src/renderer/src/components/terminal-pane/pty-transport/delivery/pty-dispatcher-delivery-interest.test.ts: registers interest on the first sidecar and releases on the last unsubscribe; releases sidecar interest only once for repeated unsubscribes; does not let an eager pre-mount buffer defeat hidden delivery gating; keeps eager buffers outside delivery interest when the PTY exits before mount; lets a sidecar exclusively own interest while an eager buffer overlaps. src/renderer/src/components/terminal-pane/pty-transport/delivery/pty-dispatcher-push-reattach.test.ts: blackholed delivery reproduces the wedge: no handler dispatch, no ACK ever. src/renderer/src/components/terminal-pane/pty-transport/ipc/ipc-pty-opencode-model.test.ts: sends preferences to the execution host and attaches only to its returned PTY; preserves one operation ID across a lost reply and reconnect; preserves ordinary raw startup and captured-session attachment. src/renderer/src/components/terminal-pane/pty-transport/output/pty-side-effect-pending-census.test.ts: sums registered gauges and drops disposed ones; ignores a repeated dispose instead of dropping another gauge; stops tracking new gauges past the cap so a missed dispose cannot grow it; tracks a live output processor queue through enqueue, drain, and dispose; excludes cleared consumed entries from retention after a bounded drain. src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-prebind-discard-notice.test.ts: says nothing once the pane is closed. src/renderer/src/components/terminal-pane/render-health/terminal-webgl-atlas-recovery.test.ts: refreshes after each scheduled atlas reset; falls back to a timeout when animation frames are unavailable; continues recovery when a manager throws after scheduling. src/renderer/src/components/terminal-pane/snapshots/pty-buffer-serializer.test.ts: tracks a remounted quiet title source even when the stale owner unregisters later; preserves the renderer-parsed output sequence in the response; preserves a pending escape tail for snapshot replay; routes a host's Reset Terminal request to the pane that owns the PTY. src/renderer/src/components/terminal-pane/snapshots/saved-terminal-output-replay.test.ts: marks panes with restored scrollback for fresh-shell viewport blanking. src/renderer/src/components/terminal-pane/viewport/split-right-white-screen.test.ts: recovers a 0x0 spawn whose container is briefly unmeasurable then settles. src/renderer/src/components/terminal-pane/viewport/visibility/stale-document-visibility.test.ts: removes global listeners when the last pane unregisters. src/renderer/src/components/ui/security-surface-theme-contrast.test.tsx: keeps trust text readable on a host-owned dialog surface; applies the hostile surface to ordinary dialogs, so the case above is exercised. src/renderer/src/components/use-terminal-parking-pass.test.tsx: does not queue a render when a pass produces the same parking sets; keeps parking transitions working under StrictMode. src/renderer/src/hooks/client-workspace-view-report.test.ts: reports an away user’s screen with no presence, and renews as soon as input returns; retries a failed send, empty or not, no sooner than a refresh. src/renderer/src/hooks/ipc-events/terminal-request-ipc-bridge-placement.test.ts: is new work in a workspace with no live terminal: its first bind moves it up in Recent. src/renderer/src/hooks/use-bundled-skill-install.test.tsx: ignores a repeated click while the install runs; retries the last requested skills on this host. src/renderer/src/hooks/useGitHubSlugMetadata.assignee-reuse.test.tsx: adds seeds to a successful empty list but not to loading or failed results; retains seed projection after expiry and fetches fresh metadata for a new reader. src/renderer/src/hooks/useSettingsNavigationMetadata.language-switch.test.tsx: retranslates the settings nav labels when the UI language changes live. src/renderer/src/lib/codex-stale-pane-account-identity.test.ts: keeps the prompt when the roster read fails and every label collapses. src/renderer/src/lib/image-blob-png.test.ts: downscales a pasted image above the pixel cap; releases its temporary URL on a decode error; releases its temporary URL when PNG encoding fails or exceeds the byte cap. src/renderer/src/lib/react-commit-cascade-store-write-samples.test.ts: names callers above the frame the write came from; restores Error.stackTraceLimit when the capture throws; stops collecting after the sample cap when captureStackTrace is unavailable; caps reported changed keys from a single wide write. src/renderer/src/lib/work-items/linear/linear-issue-context-snapshot.test.ts: applies the rendered text hard cap last and reserves the final overflow marker. src/renderer/src/runtime/rpc/targets/use-worktree-runtime-target.test.ts: keeps the runtime target identity stable across unrelated store writes. src/renderer/src/runtime/source-control/git/runtime-git-branch-client.test.ts: keeps creation available on old hosts and explains the missing operation; preserves landed warnings and direct collision recovery data; propagates remote discovery failure instead of a false empty page. src/renderer/src/runtime/sync-runtime-graph/sync-runtime-graph-key-reuse.test.ts: reuses unchanged worktree tab projections when one worktree title changes; bumps only the changed worktree version when a sibling stays byte-identical. src/renderer/src/store/repos/repos-catalog-request-coordination.test.ts: shares safe-auto fork sync cooldown attempts across store instances. src/renderer/src/store/slices/tabs/hydrated-workspace-reconciliation-batch.test.ts: collapses a 193-workspace hydration to one store write; leaves the store byte-identical to the per-workspace path; re-reconciling the batched result is a no-op, as it is for the per-workspace path. src/renderer/src/store/slices/tabs/hydrated-workspace-reconciliation-fixture.ts: dedicated zero-case helper retirement; no authored test credit. src/renderer/src/store/slices/worktrees/metadata/update-worktrees-meta.test.ts: persists each batch metadata update to its workspace owner. src/renderer/src/store/slices/worktrees/metadata/worktree-meta-rename.test.ts: dismisses a rename error on its capable remote owner; clears pending first-agent rename when the title is updated. src/renderer/src/store/terminals/store-sleep-shutdown-order.test.ts: orders capture, guards, handlers, main's stop, commit, cleanup, and exit settlement. src/renderer/src/store/terminals/terminal-first-tab-publication.test.ts: publishes the first terminal and root tab group atomically; publishes the first terminal and root tab group atomically, then marks it starting in main. src/renderer/src/store/terminals/workspace-terminal-reconnect-hydration.test.ts: restores daemon terminal wake hints before declaring the session ready. src/renderer/src/web/web-runtime-connection-auth-framing.test.ts: keeps legacy framing for sockets that cannot absorb a refused attempt. src/shared/terminal/appearance/terminal-font-weight-windows-regular.test.ts: moves the old Windows default to Regular once and queues the notice; keeps custom weights and fresh profiles without a notice; leaves macOS and Linux profiles untouched. tests/e2e/claude-agent-row-view.unit.test.ts: updates a mounted compact workflow description through the subscribed production parent; bounds upstream active-SSH visits for $population records, fallback=$fallback; derives %s destinations without requiring type or a workflow child ID; terminates a $path parent $status with no third request or intermediate toast; does not recover a first child %s. tests/e2e/cross-version-wire/agent-permission-legacy-spawn.unit.test.ts: keeps exact legacy ordering when structured creation cannot express afterTabId; can create a terminal without selecting the target worktree; preserves the legacy fresh-agent path when host authority is unavailable. tests/e2e/paired-instant-browser-tab.spec.ts: shows a paired browser tab before its create RPC resolves, then keeps it as one tab; adopts a paired browser tab without rebuilding its chrome. tests/e2e/session-tabs-decorative-title-fanout.unit.test.ts: bounds host publication and renderer application across remote worktrees; globally caps decorative freshness while every working status stays fresh. tests/e2e/session-tabs/mobile-session-active-performance.unit.test.ts: bounds saved-order publication selection at N and 10N without rematerializing navigation; bounds admitted placement reconciliation during supported reads at N and 10N; bounds selection input visits at 50 and 500 rows, including unavailable-target projection; keeps publication scoped to a changed worktree and shared by listeners at N and 10N; bounds preserved-chat merge and runtime-owned hydration at N and 10N (legacy=%s); bounds actual native retirement and shared delivery work at N and 10N. tests/e2e/session-tabs/mobile-session-active-publication.unit.test.ts: does not substitute a preserved row for an unavailable explicit legacy selection; resolves main-layout group selection %s (omitted group=%s). tests/e2e/structured-agent-session-launch-admission.unit.test.ts: opens a direct caller's terminal with the caller's own CLI arguments on a decline; opens the chat on the server that admitted it; opens only the terminal when this machine declines, without a notice; opens a new agent tab's terminal with the typed prompt when the caller names none; opens the chat, not a terminal, when this machine cannot resolve a new workspace yet; keeps its terminal fallback, as before. tests/e2e/structured-attention-prompt-relay.unit.test.ts: settles once per relayed prompt, not on every later status update. tests/e2e/structured-attention-read-retirement.unit.test.ts: Mark read on a chat never opened withdraws the prompt alert it surfaced; Mark read on a chat hidden since an earlier view covers the newer prompt it surfaced; reads a newly accepted visible prompt without sending an RPC for text-only updates. tests/e2e/structured-chat-tab-lifetime-cost.unit.test.ts: counts real raw stores, unrelated window publications and captured cleanup at %i chats; counts complete supplied-member sweep and separate final retirement at S=%i N=%i. tests/e2e/structured-session-workspace-authority.unit.test.ts: bounds actual target reveal across %i repo roots and metadata aliases. tests/e2e/terminal-agent-pane-launch.unit.test.ts: starts the host's planned command once, at the pane's size, and says it started; plans a pane whose prompt the window submits without the startup update prompt; preserves caller shell-ready delivery only for a connection-scoped adapter ($connectionId); starts no shell on plan failure (original failure: $returnLaunchFailure); cannot confirm a start whose provider answer was lost. tests/e2e/terminal-parked-memory.spec.ts: releases renderer terminal memory when hidden tabs park. tests/e2e/terminal-shortcuts.spec.ts: @headful Codex-like background output stays visible without disabling WebGL in auto mode; Shift with Russian layout text reaches the PTY as Cyrillic under kitty keyboard reporting. tests/e2e/terminal-split-activation-latency.spec.ts: records attributed CWD, spawn, bind, fixture-ready, input, and echo phases. tests/e2e/terminal-split-activation-latency-phases.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-split-activation-latency-main-probe.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-split-activation-latency-artifact.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-split-activation-latency-report.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-tab-switch-visual-restore.spec.ts: restores skipped hidden Grok output on light tab resume. tests/e2e/web-runtime-session-terminal-host-authority.unit.test.ts: delivers independent %s effort through the real prepared web callback; keeps same-ID local and runtime worktrees on the selected runtime owner; creates paired web agents through host authority so activation is mirrored; delivers generated continuation context after host-authoritative creation; seeds the chat composer for a draft that rode in on the launch command. Current changed source bytes, wrapper registrations and reduced selectors have no fresh execution or red/green credit. Unrelated surviving receipt objects keep their original narrow scope; current literal graph reachability does not establish historical import closure or equivalent replacement coverage. Current conservative command/import graph reaches src/main/persistence/worktree-sort-order-write.test.ts, src/main/runtime/construction/orca-runtime.test.ts, src/main/runtime/orca-runtime-tests/targeting-and-resilience.spec.ts, src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-04.spec.ts, src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-03.spec.ts, src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-02.spec.ts, src/main/runtime/orca-runtime-tests/gitlab-and-pr-bases-part-02.spec.ts, src/main/runtime/orca-runtime-tests/gitlab-and-pr-bases.spec.ts, src/main/runtime/orca-runtime-tests/worktree-setup-and-startup-part-03.spec.ts, src/main/runtime/orca-runtime-tests/worktree-setup-and-startup.spec.ts, src/main/runtime/orca-runtime-tests/terminal-sleep-and-teardown-part-02.spec.ts, src/main/runtime/orca-runtime-tests/mobile-summaries-part-03.spec.ts, src/main/runtime/orca-runtime-tests/mobile-summaries-part-02.spec.ts, src/main/runtime/orca-runtime-tests/mobile-summaries.spec.ts, src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-03.spec.ts, src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-02.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-13.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-12.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-11.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-10.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-08.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-07.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-05.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-04.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-02.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-05.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-04.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-03.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-02.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status.spec.ts, src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-07.spec.ts, src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-04.spec.ts, src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-03.spec.ts, src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-02.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-12.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-11.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-10.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-09.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-08.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-07.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-06.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-03.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness.spec.ts, src/main/runtime/orca-runtime-tests/headless-snapshots.spec.ts, src/main/runtime/orca-runtime-tests/terminal-side-effect-facts-part-02.spec.ts, src/main/runtime/orca-runtime-tests/local-worktree-creation-part-02.spec.ts, src/main/runtime/orca-runtime-tests/local-worktree-creation.spec.ts, src/main/runtime/integrations/skill/orca-runtime-skill-recovery.test.ts, src/main/runtime/orchestration/mailbox/routing/orchestration-mailbox-detached-routing.spec.ts, src/main/runtime/orchestration/mailbox/terminal-pointer/orchestration-mailbox-delivery.test.ts, src/main/runtime/orchestration/mailbox/terminal-pointer/orchestration-mailbox-pointer-cli-command.spec.ts, src/main/runtime/rpc/methods/terminal/terminal-multiplex-link-estimate.test.ts, src/main/runtime/rpc/methods/terminal/terminal-snapshot-scrollback-rows.test.ts, src/main/runtime/rpc/server/pairing/network/windows-default-route-interfaces.test.ts, src/main/runtime/rpc/server/pairing/pairing-offer-owner-cli.test.ts, src/main/runtime/rpc/terminal-multiplex-end-verdict.test.ts, src/main/runtime/session-graph/headless-runtime-graph.spec.ts, src/main/runtime/session-graph/runtime-graph-standing-contracts.test.ts, src/main/runtime/session-tabs/structured-conversation-tab-title.test.ts, src/main/runtime/session-tabs/terminal-close-observed-exit.test.ts, src/main/runtime/terminal/identity/terminal-identity-probe.test.ts, src/main/runtime/terminal/layout/paired-desktop-server-writes-layout-owner.test.ts, src/main/runtime/terminal/lifecycle/pty-lifecycle-generation-retention.test.ts, src/main/runtime/terminal/lifecycle/run-terminal-user-open.test.ts, src/main/runtime/terminal/surface/terminal-surface-replies-window.test.ts, src/main/runtime/terminal/surface/workspace-layout-projection.test.ts, src/main/runtime/terminal/view/terminal-subscribe-exit-waiter-leak.test.ts, src/main/runtime/terminal/wait/orca-runtime-exact-worker-provider-session.test.ts, src/main/runtime/terminal/wait/prime-agent-screen-readiness-transcripts.test.ts, src/main/runtime/workspaces/startup/runtime-worktree-agent-startup.test.ts, src/main/runtime/workspaces/startup/runtime-worktree-startup-readiness.test.ts, src/main/skills/skill-install-transaction.test.ts, src/main/speech/model-manager-download-resume.test.ts, src/main/ssh/config/ssh-config-parser.test.ts, src/main/ssh/config/ssh-known-hosts-source.test.ts, src/main/ssh/config/ssh-known-hosts.test.ts, src/main/ssh/connection/ssh-connect-config.test.ts, src/main/ssh/connection/ssh-connection-auth-fallback.test.ts, src/main/ssh/connection/ssh-connection-store.test.ts, src/main/ssh/connection/ssh-connection.test.ts, src/main/ssh/orcad-migration-stuck-host-profiles.test.ts, src/main/ssh/orcad-runtime-maintenance.test.ts, src/main/ssh/port-forwarding/ssh-port-forward.test.ts, src/main/ssh/ssh-channel-multiplexer.test.ts, src/main/ssh/system-transport/system-ssh-forward-process.test.ts, src/main/ssh/system-transport/system-ssh-loopback-route.test.ts, src/main/startup/browser-process-user-agent-ordering.test.ts, src/main/startup/quit/quit-path-durable-write-blocking.test.ts, src/main/stats/agent-session-transition-recorder.test.ts, src/main/text-generation/commit-message-text-generation-model-discovery.test.ts, src/main/text-generation/pull-request-context.test.ts, src/main/text-generation/source-control-local-process.test.ts, src/main/updater/linux-package/linux-package-update-recovery.test.ts, src/main/updater/releases/updater-prerelease-feed-readiness.test.ts, src/main/updater/releases/updater-release-builds.test.ts, src/main/updater/updater.build-channel-selection.test.ts, src/main/updater/updater.check-preflight.test.ts, src/main/updater/updater.prerelease-fallback.test.ts, src/main/updater/updater.publishing-window-feed.test.ts, src/main/updater/updater.quit-and-install.test.ts, src/main/window/createMainWindow-markdown-editor-focus.test.ts, src/main/window/createMainWindow-renderer-crash-recovery.test.ts, src/main/window/dashboard-popout-window.test.ts, src/main/window/dropped-path-preparation.test.ts, src/main/window/main-window-updater-deferred-setup.test.ts, src/main/worker-threads/worker-thread-request-queue.test.ts, src/main/workspace-hooks/hooks-setup-runner-script.test.ts, src/main/workspaces/catalog/repo-git-remote-identity-enrichment.test.ts, src/main/workspaces/creation/preparation/worktree-create-preparation.test.ts, src/main/workspaces/removal/local-worktree-removal-recovery.test.ts, src/main/workspaces/removal/worktree-removal-safety.test.ts, src/main/wsl/discovery/distro-catalog.test.ts, src/main/wsl/discovery/wsl-distro-holder.test.ts, src/main/wsl/filesystem/wsl-guest-drive-mountinfo.test.ts, src/main/wsl/orcad/wsl-orcad-deploy-app-update.test.ts, src/renderer/src/components/agent/agent-launch-picker-acquisition.test.ts, src/renderer/src/components/feature-tips/feature-tip-telemetry.test.ts, src/renderer/src/components/native-chat/composer/pickers/use-native-chat-side-chat-command.test.ts, src/renderer/src/components/native-chat/media/images/native-chat-image-copy.test.ts, src/renderer/src/components/native-chat/messages/actions/native-chat-fork-launch.test.ts, src/renderer/src/components/native-chat/messages/actions/native-chat-side-chat-fork-point.test.ts, src/renderer/src/components/native-chat/messages/actions/use-native-chat-fork.test.tsx, src/renderer/src/components/terminal-pane/provider-notices/agent-presence/agent-terminal-presence.test.ts, src/renderer/src/components/terminal-pane/pty-connection/emulation/pty-connection-mode2031-raw-subscription.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/input/pty-connection-deferred-input-forwarding.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/notifications/pty-connection-terminal-bell-attention.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/output-restore/pty-connection-dropped-output-snapshot-healing.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/output/pty-connection-fresh-native-follow-reset.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/spawn/unbound-pane-spawn-recovery.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-foreground-agent.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-hidden-replay.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-input-and-repaint.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-spawn-and-startup.test.ts, src/renderer/src/components/terminal-pane/pty-transport/delivery/pty-dispatcher-delivery-interest.test.ts, src/renderer/src/components/terminal-pane/pty-transport/delivery/pty-dispatcher-push-reattach.test.ts, src/renderer/src/components/terminal-pane/pty-transport/ipc/ipc-pty-opencode-model.test.ts, src/renderer/src/components/terminal-pane/pty-transport/output/pty-side-effect-pending-census.test.ts, src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-prebind-discard-notice.test.ts, src/renderer/src/components/terminal-pane/render-health/terminal-webgl-atlas-recovery.test.ts, src/renderer/src/components/terminal-pane/snapshots/pty-buffer-serializer.test.ts, src/renderer/src/components/terminal-pane/snapshots/saved-terminal-output-replay.test.ts, src/renderer/src/components/terminal-pane/viewport/split-right-white-screen.test.ts, src/renderer/src/components/terminal-pane/viewport/visibility/stale-document-visibility.test.ts, src/renderer/src/components/ui/security-surface-theme-contrast.test.tsx, src/renderer/src/components/use-terminal-parking-pass.test.tsx, src/renderer/src/hooks/client-workspace-view-report.test.ts, src/renderer/src/hooks/ipc-events/terminal-request-ipc-bridge-placement.test.ts, src/renderer/src/hooks/use-bundled-skill-install.test.tsx, src/renderer/src/hooks/useGitHubSlugMetadata.assignee-reuse.test.tsx, src/renderer/src/hooks/useSettingsNavigationMetadata.language-switch.test.tsx, src/renderer/src/lib/codex-stale-pane-account-identity.test.ts, src/renderer/src/lib/image-blob-png.test.ts, src/renderer/src/lib/react-commit-cascade-store-write-samples.test.ts, src/renderer/src/lib/work-items/linear/linear-issue-context-snapshot.test.ts, src/renderer/src/runtime/rpc/targets/use-worktree-runtime-target.test.ts, src/renderer/src/runtime/source-control/git/runtime-git-branch-client.test.ts, src/renderer/src/runtime/sync-runtime-graph/sync-runtime-graph-key-reuse.test.ts, src/renderer/src/store/repos/repos-catalog-request-coordination.test.ts, src/renderer/src/store/slices/tabs/hydrated-workspace-reconciliation-batch.test.ts, src/renderer/src/store/slices/tabs/hydrated-workspace-reconciliation-fixture.ts, src/renderer/src/store/slices/worktrees/metadata/update-worktrees-meta.test.ts, src/renderer/src/store/slices/worktrees/metadata/worktree-meta-rename.test.ts, src/renderer/src/store/terminals/store-sleep-shutdown-order.test.ts, src/renderer/src/store/terminals/terminal-first-tab-publication.test.ts, src/renderer/src/store/terminals/workspace-terminal-reconnect-hydration.test.ts, src/renderer/src/web/web-runtime-connection-auth-framing.test.ts, src/shared/terminal/appearance/terminal-font-weight-windows-regular.test.ts, tests/e2e/claude-agent-row-view.unit.test.ts, tests/e2e/cross-version-wire/agent-permission-legacy-spawn.unit.test.ts, tests/e2e/paired-instant-browser-tab.spec.ts, tests/e2e/session-tabs-decorative-title-fanout.unit.test.ts, tests/e2e/session-tabs/mobile-session-active-performance.unit.test.ts, tests/e2e/session-tabs/mobile-session-active-publication.unit.test.ts, tests/e2e/structured-agent-session-launch-admission.unit.test.ts, tests/e2e/structured-attention-prompt-relay.unit.test.ts, tests/e2e/structured-attention-read-retirement.unit.test.ts, tests/e2e/structured-chat-tab-lifetime-cost.unit.test.ts, tests/e2e/structured-session-workspace-authority.unit.test.ts, tests/e2e/terminal-agent-pane-launch.unit.test.ts, tests/e2e/terminal-parked-memory.spec.ts, tests/e2e/terminal-shortcuts.spec.ts, tests/e2e/terminal-split-activation-latency.spec.ts, tests/e2e/terminal-split-activation-latency-phases.ts, tests/e2e/terminal-split-activation-latency-main-probe.ts, tests/e2e/terminal-split-activation-latency-artifact.ts, tests/e2e/terminal-split-activation-latency-report.ts, tests/e2e/terminal-tab-switch-visual-restore.spec.ts, tests/e2e/web-runtime-session-terminal-host-authority.unit.test.ts. Unresolved local/computed edges remain explicit debt.", @@ -12329,14 +12332,14 @@ "providers": ["local", "remote-runtime", "ssh"], "coveredPlatforms": ["macos", "linux"], "coveredProviders": ["local", "remote-runtime", "ssh"], - "coverageNotes": "Audit3553/3559 current scope: delete worktrees-delete-pty-teardown.test.ts and worktrees-issue-command-overrides.test.ts entirely and remove three cases from worktrees-orphan-directory-cleanup.test.ts (6 authored/6 finite removals). Intentionally retire typed unserved-SSH create/delete admission, exact-name collision/error-token, proven-orphan no-force confirmation, standalone .git-directory refusal, and SSH-owned local-path preservation alarms. Retain only actual authorized orphan deletion and no-marker notes-file preservation; they are not equivalent replacements. The no-marker fixture resolves the default getRepos owner, so it does not prove temporary owner-repo binding. Git/Store/PTY/notifications are mocked; no physical process stop, durable metadata, real SSH delivery or native UI proof is claimed. Unchanged historical oracle, coverage, evidenceRuns and escaped receipts retain their original scopes only; no fresh execution or transferred pass validates the reduced sources. Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3460 final-v2 composes accepted AUTHOR3454/PEER3455 on actual post263 main: intentionally removes99 authored/106 potential expanded cases across16 selected drivers, including7 whole suites, retaining24 authored/26 potential cases. No equivalent replacement or fresh runtime is claimed. Broad root-unit discovery loses whole cross-client workspace-sync, terminal-preview, picker-skill-hint and close-routing suites plus reduced worker-stop, remote-owned-record, scanner, mobile-bundle and daemon-health samples. Separate E2E discovery also loses the paired mobile/desktop mirror journey and two unit drivers. Retained callback/table/outer-loop/guard bytes are exact; retained checks do not replace the removed alarms. No app, native device, mixed-version, remote or cross-platform execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The retained runtime registration driver loses three complete imported fragments: changed-working-label title facts and suppressed-frame tracked snapshots; refusing SSH GitLab MR-base lookup before local Git/glab calls; and missing folder-workspace path creation/update refusal before spawn/store mutation. These alarms are intentionally lost. Other imported fragments and separately rooted dated receipts are unchanged; they are not equivalent replacement coverage. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2090 (sources2088/2089): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2086 (sources2082/2083): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-handles-part-02.spec.ts: recovers a moved SSH pane through the tab it now sits in, not the one its lease froze; reports a plain expired orphan; does not report a pane whose only recent lease was superseded; does not report a pane whose only recent lease had its relay id recycled; picks the eligible orphan over a superseded predecessor that matches first; collects the same tabs the per-tab reader reports, in one sweep of the leases; reports no candidate when no lease names any of the worktree tabs. src/main/runtime/orca-runtime-tests/runtime-availability.spec.ts: starts unavailable with no authoritative window; reports the configured Windows terminal shell on status; reports floating workspace availability from settings on status; polls floating tabs with targeted PTY liveness and no repo/provider inventory; hydrates persisted tabs when the store cannot report repos; advertises browser screencast only when a renderer window is available; routes mobile Codex reset consumption through the account mutation coordinator; maps a definite pre-provider rejection into an authoritative current snapshot. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2056 (sources2054/2055): Intentionally retire these low-value alarms without replacement: src/renderer/src/runtime/files/watching/runtime-file-client-watch.test.ts: delivers remote file changes to listeners for the same worktree. src/main/runtime/orca-runtime.test.ts: ordinary registration cleanup removes only await import('./orca-runtime-tests/mobile-session-tabs-part-14.spec'). src/main/runtime/orca-runtime-tests/terminal-side-effect-facts-part-03.spec.ts: arms the stale-title timer for a seeded working title; arms the Command Code scrape from the noted spawn command; prefers the tracked title over a stale renderer lastTitle in the hydration seed; writes recent output into the current runtime output map; uses a later title handler replacement with the runtime receiver; calls the current injected listener with the runtime receiver; keeps optional reconciliation lazy after its callback is removed. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-14.spec.ts: lists them, so a paired client can mirror a migrated editor; closes one by retiring it from the host session; keeps an unsaved draft unless the close is forced; shares aggregate work but leaves a replacement pending promise alone; lets a targeted refresh finish while an aggregate refresh waits; refuses a close when the store lacks %s before the guard; keeps a later missing %s on the original caught failure path. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Exact retired assertionRefs, preserved as historical objects only: {\n \"file\": \"src/renderer/src/runtime/files/watching/runtime-file-client-watch.test.ts\",\n \"assertions\": [\n \"terminal watcher error and setup failure evict the shared entry before retrying callbacks run\",\n \"an old stream end cannot evict the replacement stream, which still delivers to its listener\",\n \"two same-worktree listeners share a synthetic subscription until the final stop unsubscribes and sends unwatch\"\n ]\n } Historical prior account (preserved verbatim): Audit1973: Delete the standalone mocked watcher cleanup suite and its entire sole-file prospective command. Retire the removed file-command transport-abort alias; preserve the independent cancellation suite and final-overflow/error/end cleanup callback, and13 unrelated receipt objects. Whole original changed dated receipts are retired, never narrowed or transferred. This metadata review claims no fresh execution, equivalent replacement, provider/PTY, live remote/mobile, rendered, cross-platform, soak or measured-budget result. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Partial runtime files-watch suite retains controlled Windows close-event waiting, shared-client survival, late physical-exit ownership/deadline cleanup and shutdown joining. Its combined Windows close/deadline/late-exit alias remains backed by unchanged callbacks. Retire native-error-close, pre-publication setup ownership and failed-delete rearm aliases; associated delegation, failure forwarding, removal routing and recovery examples are lost. These tests inject watch/subscription seams and do not observe an actual native watcher or physical process. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1925 (sources1922/1923): Partial runtime watcher-host suite loses mocked option/event/overflow/resubscription forwarding, supplied teardown-error propagation, aborted-signal/no-call and private ten-thousand-waiter-count inventories. Retire the 10,000-caller bound and failed-recovery-subscriber-ending assertion aliases. Keep same-root owner cancellation, pre-ready replacement, held root-close join, stale-generation recovery delivery and initial failed-root ownership mappings backed by retained real-host callbacks with injected subscription/error/physical-exit promises. No actual native watcher, process termination or resource bound was observed. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1683 removes low-value test projections, mock wiring, private-state assertions, inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained public behavior and refusal boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1645 removes low-value test projections, mock wiring, private-state assertions, inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained public behavior and refusal boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1640 removes low-value IPC, SSH, workspace and transcript-watcher tests without equivalent replacement coverage. Per-source reviews record retained public refusal, cancellation, execution-host and real-filesystem outcomes with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1634 removes low-value IPC, filesystem and runtime projections, argument inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained cancellation, ownership, authorization, filesystem and SQLite boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1627 removes low-value mocked daemon projections, scalar tables and duplicate terminal examples without equivalent replacement coverage. Per-source reviews record retained terminal content preservation, process ownership refusal and recovery boundaries together with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1623 removes low-value mocked daemon projections, scalar tables and duplicate terminal examples without equivalent replacement coverage. Per-source reviews record retained terminal content preservation, process ownership refusal and recovery boundaries together with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1516 removes package-electron-runtime-contract.test.mjs, a copied release-workflow platform and marker matrix. Its workflow wiring alarm is intentionally retired; no actual packaged watcher delivery was exercised and no equivalent replacement is claimed. Unaffected dated execution receipts retain their original scope; changed optional selectors have no fresh full-gate execution result. This audit removes the paired-web stopped-during-outage and late-ready direct-branch examples. The retained pre-ready cancellation, failed-unwatch retry and replay-continuity cases are not equivalent replacements for those ordering alarms. Current web-runtime-client.test.ts assertions below name only its retained pre-ready callback case; replay cases remain in their separate suite. Historical commands and results remain historical. Deterministic tests cover one shared healthy child per runtime process plus at most four bounded fault-quarantine children, native desktop, paired-runtime, and WSL snapshot processes sharing the same eight-physical-child reservation with typed serialized event-driven capacity recovery, including recursive re-wait when crash recovery reclaims an announced slot, one quarantine attempt per watch lifetime, generation-scoped cancellation, bounded termination with removable deadline waiters, replacement-crawl errors that cannot be blessed by a late readiness ack, pre-ready paired-web cancellation registered before an unbounded capacity wait, already-resolved and late physical-exit retry of a rejected paired-web teardown, physical desktop-install and local/SSH PTY teardown, stale-generation rejection, per-child stat bounds, final-overflow RPC delivery, and renderer eviction before terminal callbacks can retry. Destructive local and SSH removal fences both ID-derived and resolved-cwd terminal roots, including sibling-root/cwd combinations with reverse admission rollback, closes descendant relay watches before parent deletion, enumerates authoritative provider/cwd ownership, falls back to daemon spawn cwd before OSC 7, and keeps daemon/relay immediate shutdown pending until native exit or a bounded fail-closed error; a dead relay PID discovered during attach also settles concurrent shutdown before its stale entry is reaped, while Windows runtime watcher deadlines retain one removable physical-close waiter, treat Node's error-close path as positive physical-exit proof, and clear root ownership on late close. The relay policy test maps a standard repository's base, Git common directory, and worktree roots to one healthy supervisor, then proves a shared-child failure recovers those roots in separate quarantine supervisors with overflow and resumed events. Abortable capacity, quarantine, runtime-root, relay, SSH, and relay pre-install setup waits attach one reaction to each shared promise and explicitly remove 10,000 cancelled caller closures while one anchor remains. Built-entry macOS harnesses kill the desktop/runtime and SSH relay watcher children, then require automatic resubscription and later events while the host process and relay PTY survive. SSH bundles the child boundary beside relay.js, requires both artifacts for install completeness, and preserves registration-owned same-root cancellation. The test-value audit corrects the active renderer watch command/file reference to the existing runtime-file-client-watch.test.ts, whose three cases are retained unchanged. They observe synthetic subscription retry/fanout and unsubscribe/unwatch calls, not actual preload/main bookkeeping or physical watcher-process cleanup. The separate read/preview client suite loses forwarding, supplied-error and negative-only examples, retaining binary detection and Unicode SVG decoding. Historical evidence commands/results are unchanged. Test-value audit1293 removes src/main/providers/local-pty/local-pty-provider-session-inventory.test.ts (2 authored cases). Their isolated assertions no longer run; surviving tests are not claimed as equivalent coverage. Fresh execution against the candidate SHA is required. Audit1312 removes low-value imported specs from aggregate test drivers; old whole-driver execution receipts are archived. Retained registrations are unchanged. No equivalent replacement coverage is claimed. Audit1383: src/main/runtime/orca-runtime-tests/worktree-native-chat-setup.spec.ts reduced. Lose command-list/receipt/default/background activation matrices whose provider spawn returns canned IDs. Retained test proves actual tab allocation separation, not terminal startup. No equivalent replacement coverage is claimed. Existing exact unaffected receipts and their platform/provider/budget claims remain scoped to unchanged selectors. Changed optional selectors have no new execution result. Audit1401: src/main/runtime/orca-runtime-tests/mobile-session-tabs-duplicate-pty-repair.spec.ts removed. Loses listing-on-already-repaired-data alarm; it never established runtime repair invocation. src/main/runtime/orca-runtime-tests/mobile-session-tabs-split-agent-tab.spec.ts removed. Loses actual mobile split sibling shell versus recorded-agent resume command alarms; no equivalent coverage asserted. src/main/runtime/orca-runtime-tests/worktree-native-chat-setup.spec.ts removed. Loses separate setup/startup tab result in that manually assembled path. src/main/runtime/orca-runtime.test.ts reduced. Driver registration/import of a whole-deleted selected spec is removed; directly authored cases remain unchanged. No extra authored removal credit. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result. Audit1412: src/main/workspaces/removal/watcher-removal-gate.test.ts reduced. Loses exact removal-error compatibility with renderer predicate under shared message constants. Kept public refusal outcomes do not establish that renderer classification. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result.", + "coverageNotes": "Audit3563 scope: retire the whole Kiro OSC/hook identity suite and the OpenCode2 identity-only case; reduce Codex cached startup child-retirement and missing-session-ID tables to one case each (2 authored/4 finite removals). Intentionally lose Kiro alias/session continuity, OpenCode2 completed OSC identity continuity, cached replacement without relay-generation restart, and missing-previous-ID context-preservation alarms. Remaining provider, remote compatibility, persistence and cross-session pollution cases are not equivalent replacements. Also remove the private Codex roster-size assertion while retaining public child-absence checks. Historical evidence retains its original source scope; no new execution or transferred pass validates these reductions. Audit3553/3559 current scope: delete worktrees-delete-pty-teardown.test.ts and worktrees-issue-command-overrides.test.ts entirely and remove three cases from worktrees-orphan-directory-cleanup.test.ts (6 authored/6 finite removals). Intentionally retire typed unserved-SSH create/delete admission, exact-name collision/error-token, proven-orphan no-force confirmation, standalone .git-directory refusal, and SSH-owned local-path preservation alarms. Retain only actual authorized orphan deletion and no-marker notes-file preservation; they are not equivalent replacements. The no-marker fixture resolves the default getRepos owner, so it does not prove temporary owner-repo binding. Git/Store/PTY/notifications are mocked; no physical process stop, durable metadata, real SSH delivery or native UI proof is claimed. Unchanged historical oracle, coverage, evidenceRuns and escaped receipts retain their original scopes only; no fresh execution or transferred pass validates the reduced sources. Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3460 final-v2 composes accepted AUTHOR3454/PEER3455 on actual post263 main: intentionally removes99 authored/106 potential expanded cases across16 selected drivers, including7 whole suites, retaining24 authored/26 potential cases. No equivalent replacement or fresh runtime is claimed. Broad root-unit discovery loses whole cross-client workspace-sync, terminal-preview, picker-skill-hint and close-routing suites plus reduced worker-stop, remote-owned-record, scanner, mobile-bundle and daemon-health samples. Separate E2E discovery also loses the paired mobile/desktop mirror journey and two unit drivers. Retained callback/table/outer-loop/guard bytes are exact; retained checks do not replace the removed alarms. No app, native device, mixed-version, remote or cross-platform execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The retained runtime registration driver loses three complete imported fragments: changed-working-label title facts and suppressed-frame tracked snapshots; refusing SSH GitLab MR-base lookup before local Git/glab calls; and missing folder-workspace path creation/update refusal before spawn/store mutation. These alarms are intentionally lost. Other imported fragments and separately rooted dated receipts are unchanged; they are not equivalent replacement coverage. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2090 (sources2088/2089): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2086 (sources2082/2083): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-handles-part-02.spec.ts: recovers a moved SSH pane through the tab it now sits in, not the one its lease froze; reports a plain expired orphan; does not report a pane whose only recent lease was superseded; does not report a pane whose only recent lease had its relay id recycled; picks the eligible orphan over a superseded predecessor that matches first; collects the same tabs the per-tab reader reports, in one sweep of the leases; reports no candidate when no lease names any of the worktree tabs. src/main/runtime/orca-runtime-tests/runtime-availability.spec.ts: starts unavailable with no authoritative window; reports the configured Windows terminal shell on status; reports floating workspace availability from settings on status; polls floating tabs with targeted PTY liveness and no repo/provider inventory; hydrates persisted tabs when the store cannot report repos; advertises browser screencast only when a renderer window is available; routes mobile Codex reset consumption through the account mutation coordinator; maps a definite pre-provider rejection into an authoritative current snapshot. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2056 (sources2054/2055): Intentionally retire these low-value alarms without replacement: src/renderer/src/runtime/files/watching/runtime-file-client-watch.test.ts: delivers remote file changes to listeners for the same worktree. src/main/runtime/orca-runtime.test.ts: ordinary registration cleanup removes only await import('./orca-runtime-tests/mobile-session-tabs-part-14.spec'). src/main/runtime/orca-runtime-tests/terminal-side-effect-facts-part-03.spec.ts: arms the stale-title timer for a seeded working title; arms the Command Code scrape from the noted spawn command; prefers the tracked title over a stale renderer lastTitle in the hydration seed; writes recent output into the current runtime output map; uses a later title handler replacement with the runtime receiver; calls the current injected listener with the runtime receiver; keeps optional reconciliation lazy after its callback is removed. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-14.spec.ts: lists them, so a paired client can mirror a migrated editor; closes one by retiring it from the host session; keeps an unsaved draft unless the close is forced; shares aggregate work but leaves a replacement pending promise alone; lets a targeted refresh finish while an aggregate refresh waits; refuses a close when the store lacks %s before the guard; keeps a later missing %s on the original caught failure path. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Exact retired assertionRefs, preserved as historical objects only: {\n \"file\": \"src/renderer/src/runtime/files/watching/runtime-file-client-watch.test.ts\",\n \"assertions\": [\n \"terminal watcher error and setup failure evict the shared entry before retrying callbacks run\",\n \"an old stream end cannot evict the replacement stream, which still delivers to its listener\",\n \"two same-worktree listeners share a synthetic subscription until the final stop unsubscribes and sends unwatch\"\n ]\n } Historical prior account (preserved verbatim): Audit1973: Delete the standalone mocked watcher cleanup suite and its entire sole-file prospective command. Retire the removed file-command transport-abort alias; preserve the independent cancellation suite and final-overflow/error/end cleanup callback, and13 unrelated receipt objects. Whole original changed dated receipts are retired, never narrowed or transferred. This metadata review claims no fresh execution, equivalent replacement, provider/PTY, live remote/mobile, rendered, cross-platform, soak or measured-budget result. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Partial runtime files-watch suite retains controlled Windows close-event waiting, shared-client survival, late physical-exit ownership/deadline cleanup and shutdown joining. Its combined Windows close/deadline/late-exit alias remains backed by unchanged callbacks. Retire native-error-close, pre-publication setup ownership and failed-delete rearm aliases; associated delegation, failure forwarding, removal routing and recovery examples are lost. These tests inject watch/subscription seams and do not observe an actual native watcher or physical process. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1925 (sources1922/1923): Partial runtime watcher-host suite loses mocked option/event/overflow/resubscription forwarding, supplied teardown-error propagation, aborted-signal/no-call and private ten-thousand-waiter-count inventories. Retire the 10,000-caller bound and failed-recovery-subscriber-ending assertion aliases. Keep same-root owner cancellation, pre-ready replacement, held root-close join, stale-generation recovery delivery and initial failed-root ownership mappings backed by retained real-host callbacks with injected subscription/error/physical-exit promises. No actual native watcher, process termination or resource bound was observed. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1683 removes low-value test projections, mock wiring, private-state assertions, inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained public behavior and refusal boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1645 removes low-value test projections, mock wiring, private-state assertions, inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained public behavior and refusal boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1640 removes low-value IPC, SSH, workspace and transcript-watcher tests without equivalent replacement coverage. Per-source reviews record retained public refusal, cancellation, execution-host and real-filesystem outcomes with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1634 removes low-value IPC, filesystem and runtime projections, argument inventories and duplicate examples without equivalent replacement coverage. Per-source reviews record retained cancellation, ownership, authorization, filesystem and SQLite boundaries with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1627 removes low-value mocked daemon projections, scalar tables and duplicate terminal examples without equivalent replacement coverage. Per-source reviews record retained terminal content preservation, process ownership refusal and recovery boundaries together with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1623 removes low-value mocked daemon projections, scalar tables and duplicate terminal examples without equivalent replacement coverage. Per-source reviews record retained terminal content preservation, process ownership refusal and recovery boundaries together with injected/native/remote limits. Affected historical receipts are retired; gates without active current evidence are downgraded. Historical scope before this audit: Audit1516 removes package-electron-runtime-contract.test.mjs, a copied release-workflow platform and marker matrix. Its workflow wiring alarm is intentionally retired; no actual packaged watcher delivery was exercised and no equivalent replacement is claimed. Unaffected dated execution receipts retain their original scope; changed optional selectors have no fresh full-gate execution result. This audit removes the paired-web stopped-during-outage and late-ready direct-branch examples. The retained pre-ready cancellation, failed-unwatch retry and replay-continuity cases are not equivalent replacements for those ordering alarms. Current web-runtime-client.test.ts assertions below name only its retained pre-ready callback case; replay cases remain in their separate suite. Historical commands and results remain historical. Deterministic tests cover one shared healthy child per runtime process plus at most four bounded fault-quarantine children, native desktop, paired-runtime, and WSL snapshot processes sharing the same eight-physical-child reservation with typed serialized event-driven capacity recovery, including recursive re-wait when crash recovery reclaims an announced slot, one quarantine attempt per watch lifetime, generation-scoped cancellation, bounded termination with removable deadline waiters, replacement-crawl errors that cannot be blessed by a late readiness ack, pre-ready paired-web cancellation registered before an unbounded capacity wait, already-resolved and late physical-exit retry of a rejected paired-web teardown, physical desktop-install and local/SSH PTY teardown, stale-generation rejection, per-child stat bounds, final-overflow RPC delivery, and renderer eviction before terminal callbacks can retry. Destructive local and SSH removal fences both ID-derived and resolved-cwd terminal roots, including sibling-root/cwd combinations with reverse admission rollback, closes descendant relay watches before parent deletion, enumerates authoritative provider/cwd ownership, falls back to daemon spawn cwd before OSC 7, and keeps daemon/relay immediate shutdown pending until native exit or a bounded fail-closed error; a dead relay PID discovered during attach also settles concurrent shutdown before its stale entry is reaped, while Windows runtime watcher deadlines retain one removable physical-close waiter, treat Node's error-close path as positive physical-exit proof, and clear root ownership on late close. The relay policy test maps a standard repository's base, Git common directory, and worktree roots to one healthy supervisor, then proves a shared-child failure recovers those roots in separate quarantine supervisors with overflow and resumed events. Abortable capacity, quarantine, runtime-root, relay, SSH, and relay pre-install setup waits attach one reaction to each shared promise and explicitly remove 10,000 cancelled caller closures while one anchor remains. Built-entry macOS harnesses kill the desktop/runtime and SSH relay watcher children, then require automatic resubscription and later events while the host process and relay PTY survive. SSH bundles the child boundary beside relay.js, requires both artifacts for install completeness, and preserves registration-owned same-root cancellation. The test-value audit corrects the active renderer watch command/file reference to the existing runtime-file-client-watch.test.ts, whose three cases are retained unchanged. They observe synthetic subscription retry/fanout and unsubscribe/unwatch calls, not actual preload/main bookkeeping or physical watcher-process cleanup. The separate read/preview client suite loses forwarding, supplied-error and negative-only examples, retaining binary detection and Unicode SVG decoding. Historical evidence commands/results are unchanged. Test-value audit1293 removes src/main/providers/local-pty/local-pty-provider-session-inventory.test.ts (2 authored cases). Their isolated assertions no longer run; surviving tests are not claimed as equivalent coverage. Fresh execution against the candidate SHA is required. Audit1312 removes low-value imported specs from aggregate test drivers; old whole-driver execution receipts are archived. Retained registrations are unchanged. No equivalent replacement coverage is claimed. Audit1383: src/main/runtime/orca-runtime-tests/worktree-native-chat-setup.spec.ts reduced. Lose command-list/receipt/default/background activation matrices whose provider spawn returns canned IDs. Retained test proves actual tab allocation separation, not terminal startup. No equivalent replacement coverage is claimed. Existing exact unaffected receipts and their platform/provider/budget claims remain scoped to unchanged selectors. Changed optional selectors have no new execution result. Audit1401: src/main/runtime/orca-runtime-tests/mobile-session-tabs-duplicate-pty-repair.spec.ts removed. Loses listing-on-already-repaired-data alarm; it never established runtime repair invocation. src/main/runtime/orca-runtime-tests/mobile-session-tabs-split-agent-tab.spec.ts removed. Loses actual mobile split sibling shell versus recorded-agent resume command alarms; no equivalent coverage asserted. src/main/runtime/orca-runtime-tests/worktree-native-chat-setup.spec.ts removed. Loses separate setup/startup tab result in that manually assembled path. src/main/runtime/orca-runtime.test.ts reduced. Driver registration/import of a whole-deleted selected spec is removed; directly authored cases remain unchanged. No extra authored removal credit. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result. Audit1412: src/main/workspaces/removal/watcher-removal-gate.test.ts reduced. Loses exact removal-error compatibility with renderer predicate under shared message constants. Kept public refusal outcomes do not establish that renderer classification. No equivalent replacement coverage is claimed. Existing exact unaffected receipts remain unchanged and apply only to their unchanged selectors. Changed optional selectors have no new execution result.", "motivatingLinks": [ "https://github.com/stablyai/orca/issues/5308", "https://github.com/stablyai/orca/issues/7547", "https://github.com/stablyai/orca/issues/8212" ], "invariant": "Local runtime and SSH relay roots share one watcher child per host process while healthy and use at most four quarantine children after shard failure; native desktop, paired-runtime, and direct WSL snapshot children consume the same process-wide eight-child reservation and retry transient exhaustion without permanently disabling a root. Cancellation remains owned by its child generation and settles only after acknowledgement or bounded physical exit; paired-web retains cleanup ownership and can cancel before readiness even while capacity is exhausted, retries a failed unwatch after already-resolved or late physical exit without duplicating stream termination, replacements never overlap a still-retiring generation, and aborted or timed-out callers detach immediately even while queued behind same-root relay teardown. Destructive local and SSH cleanup holds one containment-aware watcher/terminal removal fence through descendant watcher and PTY physical exit plus parent Git/filesystem deletion; daemon and relay kill requests retain ownership until the actual subprocess exit callback or an equivalent positive dead-PID proof and propagate rejected signals or a bounded failure otherwise. A crash-fused shared shard moves affected roots into the bounded quarantine pool, a second fused quarantine ends those watch lifetimes, interrupted setup emits overflow after recovery, metadata work stays bounded, and the SSH relay, PTYs, Git, and other filesystem operations never load the watcher native addon or share its fault boundary.", - "oracle": "Audit3553/3559 current scope: delete worktrees-delete-pty-teardown.test.ts and worktrees-issue-command-overrides.test.ts entirely and remove three cases from worktrees-orphan-directory-cleanup.test.ts (6 authored/6 finite removals). Intentionally retire typed unserved-SSH create/delete admission, exact-name collision/error-token, proven-orphan no-force confirmation, standalone .git-directory refusal, and SSH-owned local-path preservation alarms. Retain only actual authorized orphan deletion and no-marker notes-file preservation; they are not equivalent replacements. The no-marker fixture resolves the default getRepos owner, so it does not prove temporary owner-repo binding. Git/Store/PTY/notifications are mocked; no physical process stop, durable metadata, real SSH delivery or native UI proof is claimed. Unchanged historical oracle, coverage, evidenceRuns and escaped receipts retain their original scopes only; no fresh execution or transferred pass validates the reduced sources. Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3460 final-v2 composes accepted AUTHOR3454/PEER3455 on actual post263 main: intentionally removes99 authored/106 potential expanded cases across16 selected drivers, including7 whole suites, retaining24 authored/26 potential cases. No equivalent replacement or fresh runtime is claimed. Broad root-unit discovery loses whole cross-client workspace-sync, terminal-preview, picker-skill-hint and close-routing suites plus reduced worker-stop, remote-owned-record, scanner, mobile-bundle and daemon-health samples. Separate E2E discovery also loses the paired mobile/desktop mirror journey and two unit drivers. Retained callback/table/outer-loop/guard bytes are exact; retained checks do not replace the removed alarms. No app, native device, mixed-version, remote or cross-platform execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The retained runtime registration driver loses three complete imported fragments: changed-working-label title facts and suppressed-frame tracked snapshots; refusing SSH GitLab MR-base lookup before local Git/glab calls; and missing folder-workspace path creation/update refusal before spawn/store mutation. These alarms are intentionally lost. Other imported fragments and separately rooted dated receipts are unchanged; they are not equivalent replacement coverage. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2090 (sources2088/2089): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2086 (sources2082/2083): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-handles-part-02.spec.ts: recovers a moved SSH pane through the tab it now sits in, not the one its lease froze; reports a plain expired orphan; does not report a pane whose only recent lease was superseded; does not report a pane whose only recent lease had its relay id recycled; picks the eligible orphan over a superseded predecessor that matches first; collects the same tabs the per-tab reader reports, in one sweep of the leases; reports no candidate when no lease names any of the worktree tabs. src/main/runtime/orca-runtime-tests/runtime-availability.spec.ts: starts unavailable with no authoritative window; reports the configured Windows terminal shell on status; reports floating workspace availability from settings on status; polls floating tabs with targeted PTY liveness and no repo/provider inventory; hydrates persisted tabs when the store cannot report repos; advertises browser screencast only when a renderer window is available; routes mobile Codex reset consumption through the account mutation coordinator; maps a definite pre-provider rejection into an authoritative current snapshot. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2056 (sources2054/2055): Intentionally retire these low-value alarms without replacement: src/renderer/src/runtime/files/watching/runtime-file-client-watch.test.ts: delivers remote file changes to listeners for the same worktree. src/main/runtime/orca-runtime.test.ts: ordinary registration cleanup removes only await import('./orca-runtime-tests/mobile-session-tabs-part-14.spec'). src/main/runtime/orca-runtime-tests/terminal-side-effect-facts-part-03.spec.ts: arms the stale-title timer for a seeded working title; arms the Command Code scrape from the noted spawn command; prefers the tracked title over a stale renderer lastTitle in the hydration seed; writes recent output into the current runtime output map; uses a later title handler replacement with the runtime receiver; calls the current injected listener with the runtime receiver; keeps optional reconciliation lazy after its callback is removed. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-14.spec.ts: lists them, so a paired client can mirror a migrated editor; closes one by retiring it from the host session; keeps an unsaved draft unless the close is forced; shares aggregate work but leaves a replacement pending promise alone; lets a targeted refresh finish while an aggregate refresh waits; refuses a close when the store lacks %s before the guard; keeps a later missing %s on the original caught failure path. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Exact retired assertionRefs, preserved as historical objects only: {\n \"file\": \"src/renderer/src/runtime/files/watching/runtime-file-client-watch.test.ts\",\n \"assertions\": [\n \"terminal watcher error and setup failure evict the shared entry before retrying callbacks run\",\n \"an old stream end cannot evict the replacement stream, which still delivers to its listener\",\n \"two same-worktree listeners share a synthetic subscription until the final stop unsubscribes and sends unwatch\"\n ]\n } Historical prior account (preserved verbatim): Audit1973: Delete the standalone mocked watcher cleanup suite and its entire sole-file prospective command. Retire the removed file-command transport-abort alias; preserve the independent cancellation suite and final-overflow/error/end cleanup callback, and13 unrelated receipt objects. Whole original changed dated receipts are retired, never narrowed or transferred. This metadata review claims no fresh execution, equivalent replacement, provider/PTY, live remote/mobile, rendered, cross-platform, soak or measured-budget result. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Partial runtime files-watch suite retains controlled Windows close-event waiting, shared-client survival, late physical-exit ownership/deadline cleanup and shutdown joining. Its combined Windows close/deadline/late-exit alias remains backed by unchanged callbacks. Retire native-error-close, pre-publication setup ownership and failed-delete rearm aliases; associated delegation, failure forwarding, removal routing and recovery examples are lost. These tests inject watch/subscription seams and do not observe an actual native watcher or physical process. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1925 (sources1922/1923): Partial runtime watcher-host suite loses mocked option/event/overflow/resubscription forwarding, supplied teardown-error propagation, aborted-signal/no-call and private ten-thousand-waiter-count inventories. Retire the 10,000-caller bound and failed-recovery-subscriber-ending assertion aliases. Keep same-root owner cancellation, pre-ready replacement, held root-close join, stale-generation recovery delivery and initial failed-root ownership mappings backed by retained real-host callbacks with injected subscription/error/physical-exit promises. No actual native watcher, process termination or resource bound was observed. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1683 retires changed-source assertion and execution scope without equivalent replacement coverage. Existing unchanged-selector receipts remain scoped to their original sources. Historical prior-source oracle: Existing exact unchanged-selector receipts remain the only current execution evidence. Changed source selectors have no fresh execution result. Historical prior-source oracle account: Existing exact unchanged-selector receipts remain the only current execution evidence. Changed source selectors have no fresh execution result. Historical prior-source oracle account: Fill the global watcher-child reservation, require native desktop, paired-runtime, and WSL setup to retain typed transient capacity and retry after one serialized release, then cancel pending desktop and paired-client watches and require no retry. Hold one live anchor while aborting 10,000 callers at the capacity, quarantine-assignment, runtime-root, relay-setup, and SSH-registration layers; require exactly one explicit waiter and one base-promise reaction to remain, then zero after settlement. Hold an old relay unsubscribe, queue one same-root setup leader and one anchor, then abort 10,000 joiners; require every cancellation to settle before teardown releases, exactly one anchor waiter, and only one later physical install. Advance the former paired-web callback deadline while capacity remains full, require cancellation ownership to stay live, publish the server token before ready, and require files.unwatch to abort setup before a later slot release can install a native root. Reject the first ready-watch unwatch with a physical-exit promise, resolve that exit later, and require a fresh physical unwatch attempt, registry release, and exactly one stream end. Fill one healthy shared child with multiple roots, map a standard SSH repository's base, Git common directory, and worktree roots to one relay supervisor, then inject IPC disconnect-before-exit, queued and active cancellation, operations during termination, kill-without-exit, repeated unkillable generations, initial and recovery timeouts, and shard failure. During local and SSH deletion, attempt a late terminal spawn whose worktree identity and cwd name different roots, enumerate a non-prefix remote PTY by authoritative worktree ownership, withhold daemon/relay exit callbacks, make attach positively prove a concurrent-shutdown PID dead, and make the provider unavailable; require every acquired admission to roll back on rejection, dead-PID proof to settle physical exit before reaping, timeout to retain ownership, and Git/filesystem deletion never to start. Require bounded watcher process count, physical-exit settlement, healthy-root recovery, stale-generation rejection, overflow plus later events, and fail-closed destructive cleanup. Verify relay build hashes and remote install probes require both relay.js and relay-watcher.js. Launch the built relay with a real PTY and Parcel watch, kill only the watcher child, and require the same relay PID and PTY to answer before a replacement child delivers overflow and a post-crash event.", + "oracle": "Audit3563 scope: retire the whole Kiro OSC/hook identity suite and the OpenCode2 identity-only case; reduce Codex cached startup child-retirement and missing-session-ID tables to one case each (2 authored/4 finite removals). Intentionally lose Kiro alias/session continuity, OpenCode2 completed OSC identity continuity, cached replacement without relay-generation restart, and missing-previous-ID context-preservation alarms. Remaining provider, remote compatibility, persistence and cross-session pollution cases are not equivalent replacements. Also remove the private Codex roster-size assertion while retaining public child-absence checks. Historical evidence retains its original source scope; no new execution or transferred pass validates these reductions. Audit3553/3559 current scope: delete worktrees-delete-pty-teardown.test.ts and worktrees-issue-command-overrides.test.ts entirely and remove three cases from worktrees-orphan-directory-cleanup.test.ts (6 authored/6 finite removals). Intentionally retire typed unserved-SSH create/delete admission, exact-name collision/error-token, proven-orphan no-force confirmation, standalone .git-directory refusal, and SSH-owned local-path preservation alarms. Retain only actual authorized orphan deletion and no-marker notes-file preservation; they are not equivalent replacements. The no-marker fixture resolves the default getRepos owner, so it does not prove temporary owner-repo binding. Git/Store/PTY/notifications are mocked; no physical process stop, durable metadata, real SSH delivery or native UI proof is claimed. Unchanged historical oracle, coverage, evidenceRuns and escaped receipts retain their original scopes only; no fresh execution or transferred pass validates the reduced sources. Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3460 final-v2 composes accepted AUTHOR3454/PEER3455 on actual post263 main: intentionally removes99 authored/106 potential expanded cases across16 selected drivers, including7 whole suites, retaining24 authored/26 potential cases. No equivalent replacement or fresh runtime is claimed. Broad root-unit discovery loses whole cross-client workspace-sync, terminal-preview, picker-skill-hint and close-routing suites plus reduced worker-stop, remote-owned-record, scanner, mobile-bundle and daemon-health samples. Separate E2E discovery also loses the paired mobile/desktop mirror journey and two unit drivers. Retained callback/table/outer-loop/guard bytes are exact; retained checks do not replace the removed alarms. No app, native device, mixed-version, remote or cross-platform execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The retained runtime registration driver loses three complete imported fragments: changed-working-label title facts and suppressed-frame tracked snapshots; refusing SSH GitLab MR-base lookup before local Git/glab calls; and missing folder-workspace path creation/update refusal before spawn/store mutation. These alarms are intentionally lost. Other imported fragments and separately rooted dated receipts are unchanged; they are not equivalent replacement coverage. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2090 (sources2088/2089): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2086 (sources2082/2083): Current reduced suites intentionally retire the alarms listed in knownGaps; prior dated execution and replacement claims do not validate changed source bytes or selectors. Historical prior account (preserved verbatim): Audit2076 (sources2073/2074): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-handles-part-02.spec.ts: recovers a moved SSH pane through the tab it now sits in, not the one its lease froze; reports a plain expired orphan; does not report a pane whose only recent lease was superseded; does not report a pane whose only recent lease had its relay id recycled; picks the eligible orphan over a superseded predecessor that matches first; collects the same tabs the per-tab reader reports, in one sweep of the leases; reports no candidate when no lease names any of the worktree tabs. src/main/runtime/orca-runtime-tests/runtime-availability.spec.ts: starts unavailable with no authoritative window; reports the configured Windows terminal shell on status; reports floating workspace availability from settings on status; polls floating tabs with targeted PTY liveness and no repo/provider inventory; hydrates persisted tabs when the store cannot report repos; advertises browser screencast only when a renderer window is available; routes mobile Codex reset consumption through the account mutation coordinator; maps a definite pre-provider rejection into an authoritative current snapshot. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Frozen current literal-import graph reaches these changed test bytes; unresolved local and computed edges remain debt, and current reachability is not historical execution or equivalent replacement evidence. Historical prior account (preserved verbatim): Audit2056 (sources2054/2055): Intentionally retire these low-value alarms without replacement: src/renderer/src/runtime/files/watching/runtime-file-client-watch.test.ts: delivers remote file changes to listeners for the same worktree. src/main/runtime/orca-runtime.test.ts: ordinary registration cleanup removes only await import('./orca-runtime-tests/mobile-session-tabs-part-14.spec'). src/main/runtime/orca-runtime-tests/terminal-side-effect-facts-part-03.spec.ts: arms the stale-title timer for a seeded working title; arms the Command Code scrape from the noted spawn command; prefers the tracked title over a stale renderer lastTitle in the hydration seed; writes recent output into the current runtime output map; uses a later title handler replacement with the runtime receiver; calls the current injected listener with the runtime receiver; keeps optional reconciliation lazy after its callback is removed. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-14.spec.ts: lists them, so a paired client can mirror a migrated editor; closes one by retiring it from the host session; keeps an unsaved draft unless the close is forced; shares aggregate work but leaves a replacement pending promise alone; lets a targeted refresh finish while an aggregate refresh waits; refuses a close when the store lacks %s before the guard; keeps a later missing %s on the original caught failure path. Existing unrelated evidenceRuns retain their original scope; unchanged dated receipt objects do not validate changed source bytes or reduced selectors. No fresh gate execution or historical import-closure equivalence is claimed. Exact retired assertionRefs, preserved as historical objects only: {\n \"file\": \"src/renderer/src/runtime/files/watching/runtime-file-client-watch.test.ts\",\n \"assertions\": [\n \"terminal watcher error and setup failure evict the shared entry before retrying callbacks run\",\n \"an old stream end cannot evict the replacement stream, which still delivers to its listener\",\n \"two same-worktree listeners share a synthetic subscription until the final stop unsubscribes and sends unwatch\"\n ]\n } Historical prior account (preserved verbatim): Audit1973: Delete the standalone mocked watcher cleanup suite and its entire sole-file prospective command. Retire the removed file-command transport-abort alias; preserve the independent cancellation suite and final-overflow/error/end cleanup callback, and13 unrelated receipt objects. Whole original changed dated receipts are retired, never narrowed or transferred. This metadata review claims no fresh execution, equivalent replacement, provider/PTY, live remote/mobile, rendered, cross-platform, soak or measured-budget result. Historical prior account (preserved verbatim): Audit1931 (sources1928/1929): Partial runtime files-watch suite retains controlled Windows close-event waiting, shared-client survival, late physical-exit ownership/deadline cleanup and shutdown joining. Its combined Windows close/deadline/late-exit alias remains backed by unchanged callbacks. Retire native-error-close, pre-publication setup ownership and failed-delete rearm aliases; associated delegation, failure forwarding, removal routing and recovery examples are lost. These tests inject watch/subscription seams and do not observe an actual native watcher or physical process. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1925 (sources1922/1923): Partial runtime watcher-host suite loses mocked option/event/overflow/resubscription forwarding, supplied teardown-error propagation, aborted-signal/no-call and private ten-thousand-waiter-count inventories. Retire the 10,000-caller bound and failed-recovery-subscriber-ending assertion aliases. Keep same-root owner cancellation, pre-ready replacement, held root-close join, stale-generation recovery delivery and initial failed-root ownership mappings backed by retained real-host callbacks with injected subscription/error/physical-exit promises. No actual native watcher, process termination or resource bound was observed. Preserve meaningful remaining selectors/assertions and unchanged dated receipts within their original scope. Reduced selectors have no fresh gate execution evidence in this metadata packet. No equivalent replacement, physical watcher/PTY/provider, live SSH/remote/mobile, rendered, cross-platform, soak or measured-budget result is claimed. Historical prior account (preserved verbatim): Audit1683 retires changed-source assertion and execution scope without equivalent replacement coverage. Existing unchanged-selector receipts remain scoped to their original sources. Historical prior-source oracle: Existing exact unchanged-selector receipts remain the only current execution evidence. Changed source selectors have no fresh execution result. Historical prior-source oracle account: Existing exact unchanged-selector receipts remain the only current execution evidence. Changed source selectors have no fresh execution result. Historical prior-source oracle account: Fill the global watcher-child reservation, require native desktop, paired-runtime, and WSL setup to retain typed transient capacity and retry after one serialized release, then cancel pending desktop and paired-client watches and require no retry. Hold one live anchor while aborting 10,000 callers at the capacity, quarantine-assignment, runtime-root, relay-setup, and SSH-registration layers; require exactly one explicit waiter and one base-promise reaction to remain, then zero after settlement. Hold an old relay unsubscribe, queue one same-root setup leader and one anchor, then abort 10,000 joiners; require every cancellation to settle before teardown releases, exactly one anchor waiter, and only one later physical install. Advance the former paired-web callback deadline while capacity remains full, require cancellation ownership to stay live, publish the server token before ready, and require files.unwatch to abort setup before a later slot release can install a native root. Reject the first ready-watch unwatch with a physical-exit promise, resolve that exit later, and require a fresh physical unwatch attempt, registry release, and exactly one stream end. Fill one healthy shared child with multiple roots, map a standard SSH repository's base, Git common directory, and worktree roots to one relay supervisor, then inject IPC disconnect-before-exit, queued and active cancellation, operations during termination, kill-without-exit, repeated unkillable generations, initial and recovery timeouts, and shard failure. During local and SSH deletion, attempt a late terminal spawn whose worktree identity and cwd name different roots, enumerate a non-prefix remote PTY by authoritative worktree ownership, withhold daemon/relay exit callbacks, make attach positively prove a concurrent-shutdown PID dead, and make the provider unavailable; require every acquired admission to roll back on rejection, dead-PID proof to settle physical exit before reaping, timeout to retain ownership, and Git/filesystem deletion never to start. Require bounded watcher process count, physical-exit settlement, healthy-root recovery, stale-generation rejection, overflow plus later events, and fail-closed destructive cleanup. Verify relay build hashes and remote install probes require both relay.js and relay-watcher.js. Launch the built relay with a real PTY and Parcel watch, kill only the watcher child, and require the same relay PID and PTY to answer before a replacement child delivers overflow and a post-crash event.", "commands": [ "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/filesystem/watchers/filesystem-watcher.test.ts src/main/ipc/filesystem/watcher-process/child-lifecycle/parcel-watcher-child-registry.test.ts src/main/ipc/filesystem/watcher-process/parcel-watcher-process.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-supervisor-capacity.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-disconnect-termination.test.ts src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts src/main/ipc/worktrees-orphan-directory-cleanup.test.ts src/main/ipc/worktrees-remove-preflight.test.ts src/main/workspaces/removal/watcher-removal-gate.test.ts src/main/runtime/files/watch/orca-runtime-files-watch.test.ts src/main/runtime/workspaces/removal/worktree-teardown.test.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/methods/files/files.test.ts src/main/runtime/rpc/methods/files/files-watch-cancellation.test.ts src/renderer/src/web/web-runtime-client.test.ts src/shared/cross-platform-path.test.ts src/shared/physical-exit-tracker.test.ts src/shared/promise-settlement-waiters.test.ts src/main/ipc/pty/pty-spawn-cwd-fallback.test.ts src/main/providers/local-pty/local-pty-provider-shutdown.test.ts src/main/daemon/providers/adapter/testing/daemon-pty-adapter.test.ts src/main/daemon/pty-subprocess/testing/pty-subprocess.test.ts src/main/daemon/sessions/testing/session.test.ts src/main/daemon/terminal-host/testing/terminal-host.test.ts src/main/daemon/terminal-host/testing/terminal-host-session-reaping-leak.test.ts --reporter=dot", "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/filesystem/watchers/filesystem-watcher.test.ts src/main/ipc/filesystem/watcher-process/child-lifecycle/parcel-watcher-child-registry.test.ts src/main/ipc/filesystem/watcher-process/parcel-watcher-process.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-supervisor-capacity.test.ts src/main/ipc/filesystem/watcher-process/supervision/parcel-watcher-disconnect-termination.test.ts src/main/ipc/filesystem/watcher-process/event-delivery/parcel-watcher-event-delivery.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-local-unsubscribe.test.ts src/main/ipc/filesystem/watchers/filesystem-watcher-wsl.test.ts src/main/ipc/worktrees-orphan-directory-cleanup.test.ts src/main/ipc/worktrees-remove-preflight.test.ts src/main/workspaces/removal/watcher-removal-gate.test.ts src/main/runtime/files/watch/orca-runtime-files-watch.test.ts src/main/runtime/workspaces/removal/worktree-teardown.test.ts src/main/runtime/construction/orca-runtime.test.ts src/main/runtime/rpc/methods/files/files.test.ts src/main/runtime/rpc/methods/files/files-watch-cancellation.test.ts src/renderer/src/web/web-runtime-client.test.ts src/shared/cross-platform-path.test.ts src/shared/physical-exit-tracker.test.ts src/main/ipc/pty/pty-spawn-cwd-fallback.test.ts src/main/providers/local-pty/local-pty-provider-shutdown.test.ts src/main/daemon/providers/adapter/testing/daemon-pty-adapter.test.ts src/main/daemon/pty-subprocess/testing/pty-subprocess.test.ts src/main/daemon/sessions/testing/session.test.ts src/main/daemon/terminal-host/testing/terminal-host.test.ts src/main/daemon/terminal-host/testing/terminal-host-session-reaping-leak.test.ts --reporter=dot", @@ -12657,6 +12660,7 @@ "Add a packaged orca serve pairing scenario that proves a connected client refreshes after child replacement." ], "knownGaps": [ + "Audit3563 scope: retire the whole Kiro OSC/hook identity suite and the OpenCode2 identity-only case; reduce Codex cached startup child-retirement and missing-session-ID tables to one case each (2 authored/4 finite removals). Intentionally lose Kiro alias/session continuity, OpenCode2 completed OSC identity continuity, cached replacement without relay-generation restart, and missing-previous-ID context-preservation alarms. Remaining provider, remote compatibility, persistence and cross-session pollution cases are not equivalent replacements. Also remove the private Codex roster-size assertion while retaining public child-absence checks. Historical evidence retains its original source scope; no new execution or transferred pass validates these reductions.", "Audit3553/3559 current scope: delete worktrees-delete-pty-teardown.test.ts and worktrees-issue-command-overrides.test.ts entirely and remove three cases from worktrees-orphan-directory-cleanup.test.ts (6 authored/6 finite removals). Intentionally retire typed unserved-SSH create/delete admission, exact-name collision/error-token, proven-orphan no-force confirmation, standalone .git-directory refusal, and SSH-owned local-path preservation alarms. Retain only actual authorized orphan deletion and no-marker notes-file preservation; they are not equivalent replacements. The no-marker fixture resolves the default getRepos owner, so it does not prove temporary owner-repo binding. Git/Store/PTY/notifications are mocked; no physical process stop, durable metadata, real SSH delivery or native UI proof is claimed. Unchanged historical oracle, coverage, evidenceRuns and escaped receipts retain their original scopes only; no fresh execution or transferred pass validates the reduced sources.", "Launch direct-owner cleanup current scope: This gate’s unfiltered unit command now selects a suite reduced by 82 whole registrations / 159 potential table examples in 10 whole test Programs, including 8 registrations / 15 examples from five once-registered create-message factory closures. Lost alarms include frozen first-message/options and legacy-host fallback; distinct actions and redelivery; uncertain/refused/reloaded input custody, Retry and late proof; rich prompt/attachment upload Stop/close and handback; ordinary draft/editor/IME/StrictMode lifetime and accepted topology; provisional fallback tab placement/group/pin/order and rich worktree handoff. Four whole consent/resume-identity cases remain in two Programs; they do not replace these alarms. This entry governs current reduced-unit scope and qualifies any broader implication of unchanged oracle/coverageNotes history. Entire dated evidenceRuns, original commands/currentCommand values, counts/results and existing history retain their original scopes only; no fresh execution, transferred pass, replacement equivalence or prior upstream-removal credit is claimed.", "Audit2145 accepts removal of canned runtime metadata/guard/forwarding, fake in-memory persistence, private work-count and wrapper-duplication tests from the construction/orca-runtime.test.ts aggregate, plus the two automation run-context guards from runtime-workspace-launch-contracts.test.ts. Retained race, framing, physical completion and ownership callbacks remain byte-identical. Existing commands remain valid selectors; historical receipts stay unchanged and provide no fresh validation or replacement coverage for the reduced suites.", @@ -14010,10 +14014,10 @@ "providers": ["paired-runtime"], "coveredPlatforms": ["linux"], "coveredProviders": ["paired-runtime"], - "coverageNotes": "Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit1579 removes workflow/manifest/regex/private binary-header and diagnostic metric inventories, fixture selectors, injected forwarding/platform/signature/process examples, unproven overlap and internal churn counters without equivalent replacement coverage. Retains actual Git trust/diffs; filesystem hash/ELF-region/native discovery; child CLI module repair; compiler/lock/publication/rollback and synthetic Electron install/cache ownership; executable shell quoting/exit; controlled notary/timer cleanup; local patched extractor file/AST freshness; real Chromium/WebKit CSP/sandbox/HTTPS image/referrer/link/preservation controls under parsed Android headers. Five unchanged setup-pnpm/VM request-cleanup definitions reviewed and not rerun. GitHub, installer/archive, OS metric, client, native delegate, compressor, builder and Apple boundaries injected or absent where recorded. No actual native PTY, shipped full/native app, real Actions/cache/npm signature authority, downloaded Electron archive, real objdump/linker/native CLI/AppImage, signing/NSIS/notary/upload/updater or live freeze proof. Historical scope before this audit: Published Linux 1.4.188 desktop against current source in both directions; scheduled weekly and manually runnable. No required PR check.", + "coverageNotes": "Audit3563 scope: retire the whole Kiro OSC/hook identity suite and the OpenCode2 identity-only case; reduce Codex cached startup child-retirement and missing-session-ID tables to one case each (2 authored/4 finite removals). Intentionally lose Kiro alias/session continuity, OpenCode2 completed OSC identity continuity, cached replacement without relay-generation restart, and missing-previous-ID context-preservation alarms. Remaining provider, remote compatibility, persistence and cross-session pollution cases are not equivalent replacements. Also remove the private Codex roster-size assertion while retaining public child-absence checks. Historical evidence retains its original source scope; no new execution or transferred pass validates these reductions. Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit1579 removes workflow/manifest/regex/private binary-header and diagnostic metric inventories, fixture selectors, injected forwarding/platform/signature/process examples, unproven overlap and internal churn counters without equivalent replacement coverage. Retains actual Git trust/diffs; filesystem hash/ELF-region/native discovery; child CLI module repair; compiler/lock/publication/rollback and synthetic Electron install/cache ownership; executable shell quoting/exit; controlled notary/timer cleanup; local patched extractor file/AST freshness; real Chromium/WebKit CSP/sandbox/HTTPS image/referrer/link/preservation controls under parsed Android headers. Five unchanged setup-pnpm/VM request-cleanup definitions reviewed and not rerun. GitHub, installer/archive, OS metric, client, native delegate, compressor, builder and Apple boundaries injected or absent where recorded. No actual native PTY, shipped full/native app, real Actions/cache/npm signature authority, downloaded Electron archive, real objdump/linker/native CLI/AppImage, signing/NSIS/notary/upload/updater or live freeze proof. Historical scope before this audit: Published Linux 1.4.188 desktop against current source in both directions; scheduled weekly and manually runnable. No required PR check.", "motivatingLinks": ["https://github.com/stablyai/orca/actions/runs/34069063016"], "invariant": "A paired client and host without client-hosted browser capabilities retain server-hosted browser placement across supported version skew.", - "oracle": "Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Require both existing named browser placement scenarios to pass three times with one attempt, zero skips, zero failures, and no report errors.", + "oracle": "Audit3563 scope: retire the whole Kiro OSC/hook identity suite and the OpenCode2 identity-only case; reduce Codex cached startup child-retirement and missing-session-ID tables to one case each (2 authored/4 finite removals). Intentionally lose Kiro alias/session continuity, OpenCode2 completed OSC identity continuity, cached replacement without relay-generation restart, and missing-previous-ID context-preservation alarms. Remaining provider, remote compatibility, persistence and cross-session pollution cases are not equivalent replacements. Also remove the private Codex roster-size assertion while retaining public child-absence checks. Historical evidence retains its original source scope; no new execution or transferred pass validates these reductions. Audit3492 composes ROOT-approved75(AUTHOR3475/PEER3476 and AUTHOR3485/PEER3481, including five-field clarification and zero-credit selector3483) plus123(AUTHOR3486/PEER3487 reconciled FINAL3491, including peer case10:3 and9:0 wording clarifications). It intentionally removes198 authored registrations/232 finite literal-table potential registrations across70 changed JS/TS test sources, including45 whole suites;87 authored/90 finite potential registrations remain across95 reviewed JS/TS bodies. One mobile native-suspension manual recipe and the now-unreferenced GitLab mock fixture helper are separately removed with zero registration credit. Static literal table potential is not execution. No equivalent replacement, fresh runtime result or merge credit is claimed. Earlier135 complete dated receipts, all raw gaps and full prior98/161/history notes retain their exact scopes and bytes. The additional123 retirement accepts loss of mocked GitLab request/error/identity/classification inventories; basic Linear migration and metadata discovery; simulated OpenCode key failed-save retention/retry/log redaction; analytics identity-flush smoke that did not prove ordering; AppImage static entry guards and empty-staging cleanup; GPU/process/minidump breadcrumb/annotation shapes and weak redaction oracles; namespace/status/foreign-launcher ownership guards and persistence roundtrips; profile transfer field projections; worker hook/screen/incarnation/result admission/classification guards; and source-coupled CLI transaction fixtures. Privileged publication failure injection was a false-positive: its mkdir regex missed -m 700 and outer rejection accepted an inner assertion before any shell ran. The identity-flush case awaited identity after flush and could pass despite early flush. Retained actual negative-token cache recovery, SQL filter-before-LIMIT open approval, unreadable-root FTS preservation/degradation, latest-connect-wins persistence, worker stop-debt settlement, unmodified same-inode foreign-content transaction, explicit incomplete-pagination error, FTS own-message snippet association, filter-before-route result and bounded-search truncation honesty are distinct bounded outcomes, not replacements for retired alarms. Case9:0 checks both roots before chmod but only mounted-session/degraded-root after chmod. Entire extraction-root suite is removed; no retained extraction cleanup is claimed. Removing the manual iOS foreground-stall shell script loses its unique real native-suspension recipe; timer tests do not replace it. First-match PIDs, indefinite waiting, broad success matching and absent trap cleanup made that diagnostic brittle and able to leave a desktop stopped. The current Windows operational driver remains byte-exact, preserving already-merged removal of orcad-convert and orcad-cli-convert selectors and the retained cmd/powershell/managed CLI lanes. The package skill-sharing release selector loses only the exact already-retired skills-suite token; no extra test credit. Included approved75 retirement account follows (historical proposal labels identify provenance; final198 counts above control): Audit3488 composes ROOT-approved source49(AUTHOR3485/PEER3481, including its immutable five-field clarification) and source26(AUTHOR3475/PEER3476): intentionally removes75 authored registrations/87 finite potential registrations across35 changed sources, including25 whole suites;35 authored/36 finite potential registrations remain in the combined reviewed scopes. Literal table potential counts are not executions. This source-only reduction has no equivalent replacement, fresh runtime result or merge credit. Root unit discovery deliberately loses entire message-row filtering/chunk reconstruction, discussion-pagination rejection passthrough, Linear bearer/stack redaction, new-profile telemetry/default/existing-JSON no-op, WSL user-trigger routing, terminal-delivery refusal, native-chat worker-approval UI, followup-paste routing, live-surface tombstone refusal, activity-bar hydration, fabricated approval-caller policy, child-PATH transformation and branch-review URL/identity suites. It also loses cache roundtrip/mode/fingerprint/fallback/no-op samples; GPU field-shape privacy examples; older auth-probe failure retention; orphan SQLite-sidecar/future-schema admission; usage tie/duplicate/composite-key examples; allocation-name sanitizer/unknown-prefix examples; real worker-and-backup preflight wrapper smoke; and JSONL descriptor-loop truncation. The discussion page contained id-only discussions without notes, so its rejection did not prove a nonempty notes-prefix oracle. The preflight did launch actual worker/backup entrypoints with production internal self-validation; removing its revision/version wrapper deliberately retires that real-entry smoke alarm. The skipped-prefix JSONL keeper rewrites/grows after collecting the suffix and before final validation, not concurrently during parsing. Retained queue rejection/recovery and held/drain mechanisms, recorder nonsettling-provider deadline, corrupt authoritative-DB refusal, legacy cache split migration, merge-added-row indexing, external-frame redaction, and pinned JSONL read/verification/mutation checks are bounded concrete oracles, not equivalent replacements for retired alarms. The full cloud relay cohort/ceiling/control suite is separately removed from the relay package runner, outside root unit discovery; no cloud runtime result is claimed. The Python computer-use benchmark remains unchanged tooling with zero authored/removal credit; no native benchmark trial ran. The package skill-sharing release selector drops only the exact already-deleted skills-suite token, with zero additional removal credit. Prior dated receipts, raw gaps, and complete earlier161/98/history notes remain exactly at their original scopes; no changed-source historical receipt becomes fresh reduced-suite evidence. Included approved26 alarm account: Broad root-unit discovery loses entire ref-label string, hostile/malformed CSV-link rendering, Codex filename-discovery, mocked paged-preview postscan mutation, usage path-containment, real symlink-ancestor, Codex skip RPC, local/remote inference routing, width-preference, and synthetic workspace-document input suites. It also loses the hand-assembled sorted CSV edit example, connect-time JPEG seed and invalid AVCC guard samples, and eight-record/64KiB detector boundary cases. The symlink suite exercised actual kernel traversal; its lost ancestor alarm could omit supplemental local editor watches and leave content stale, not bypass authorization. Retained linked stream cancellation, preview-only save fence, failed-flush dirty-draft preservation, clipboard byte/edit preservation, fragmented AVCC parsing and independent quoted-CSV interpretation checks do not replace these deliberately retired alarms. The stream fixture uses an in-memory source and does not establish actual fd cleanup; editor save endpoints are synthetic, browser events are synthetic, and no native focus, actual remote delivery, real disk write, cross-platform execution or interoperability is claimed. Earlier complete dated receipts, raw known gaps and previous notes are historical evidence at their original source/selector scope, not fresh evidence for this reduced suite. Audit3438 composes ROOT-adopted3434/independent3436: intentionally removes60 authored/62 potential expanded cases across16 sources, deleting9 whole suites. Eight authored/eight potential cases remain; delete the102line unused ACP fixture and57line unused teardown helper with zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses ACP model-catalog fixtures, resume-record projections, dashboard mock payload/state cases, mobile folder/launch/session examples and dictionary pipeline rows. Selected E2E removals also retire auto-conversion and golden-core journeys and one Windows CLI conversion registration. Retain ACP reply text across a resumed subagent, both VAD checks, focused tab send ownership, async focus ownership in happy-dom, the question-skip journey, the original managed Windows CLI lifecycle journey and dictionary substitution. These checks do not replace the removed alarms. Audit3470 composes ROOT-approved AUTHOR3466/PEER3467 source proposal: intentionally removes38 authored registrations/59 finite potential expansions, including8 whole registering suites, retaining13 authored registrations/18 finite potential expansions. The sole dead shell-recipe caller cleanup has zero additional test credit. No equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses process-import boundary, Korean/Japanese locale policy, worker-reset, mobile agent-history/queued-message and median-summary suites, plus packaged mobile-manifest, worker-pool fixture, HTML preview, mobile shell and terminal samples. The environment-provider contract and its sole shell-recipe caller are deleted. Separate cross-version-wire discovery loses the large byte-budget fresh-snapshot/replay sample while retaining independent live same-revision note publication; the retained terminal stale-refresh and dirty HTML draft persistence checks do not replace removed alarms. The source4 unused start/stop fixture hooks are removed; its retained nested failed-import test remains byte-exact with its own lifecycle. Unchanged SVG preview and Git-resolution registrations stay intact. Root balanced CI excludes the cross-version directory, which has a separate job; unbalanced root discovery can include its parent. No native app/device, actual remote, released-pair interoperability or cross-platform execution result is claimed. Audit3468 accepted source proposal AUTHOR3462/PEER3463, composed privately by audit3469: removes62 authored registrations/95 finite potential expansions, including5 whole suites, retaining25 authored registrations/29 finite potential expansions. ROOT approved source value; no equivalent replacement, fresh runtime or merge credit is claimed. Broad root-unit discovery loses the profile-state worker smoke, worker-agent views and three search discovery suites, plus samples from node-server qualification, skill-bundle generation, CLI hooks/profile/recovery/offline admission, permission-hook holds and search indexing. The separate Cloud relay runner loses one relay blackbox sample; unchanged mobile native-chat and dependent-transcript drivers remain. Retained source callback/table/ancestor/fixture bytes bind to the sealed candidate. No physical-device, mixed-version, remote, cross-platform or app execution result is claimed. Audit3456 composes ROOT-adopted3449/independent3450: intentionally removes62 authored/82 potential expanded cases from16 selected drivers, including7 whole suites, while13 authored/13 potential cases remain. Delete the unused48line legacy workspace-session JSON fixture with zero test credit. These are source-only reductions without equivalent replacement or fresh runtime claims. Broad unit discovery intentionally loses whole trust-dialog observer, background-turn notifications, redirected session refusal, typed-agent identity, chat-tab lifetime-cost, model-manager and mobile diff-review controller suites, plus individual persistence, synthetic composition, pairing, manifest and subprocess samples. Retain deferred snapshot dirty-domain persistence, two direct editor ownership cases, stream counter-domain distinction, two persisted pairing identity cases, parser entrypoint asymmetry, dictation retry and process-group timeout; paired restart and packaged old-binary/corrupt-profile journeys remain in their E2E driver scope. Retained checks do not replace removed alarms. No fresh execution, native IME/microphone, real authentication, mixed-version/platform or performance result is claimed. Audit3451 composes ROOT-adopted3444/independent3445: intentionally removes101 authored registrations across16 reviewed sources, deleting7 whole suites and reducing8; one reviewed source is unchanged. Fresh main has123 authored registrations before and22 after because an upstream direct-child shell regression remains verbatim with zero removal credit. Registered removals are100+7*S and explicit-active removals99+7*S, where S is the unprobed count0..3 of admitted bash/zsh/fish shells. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses extra stream lifecycle/forwarding examples, unreadable-store/cache projections, orchestration presentation/benchmark cases, remote-owned routing matrices, mobile legacy navigation and model-catalog listing/cache examples. Whole suites retire renderer client-event projections, manually assembled structured-session reader ownership, mobile prompt presentation, renderer AI-vault resume commands, mobile Jira URL lookup, composer retention and mobile folder source-repository queries. Retained checks preserve original stream/store/orchestration/routing/navigation/catalog cases and shell behavior, including the upstream direct-child case. No real provider, native desktop or mobile acceptance is established by this reduction. Audit3446 composes ROOT-adopted3441/independent3442: intentionally removes100 authored registrations/119 potential expanded cases (118 conditional-active before platform gates) across16 sources, retaining11 authored/13 expanded cases and deleting6 whole suites. Two parent registrar import/call cleanups and the unused68line CLI recording fixture carry zero test credit. No equivalent replacement or fresh execution is claimed. Broad unit and mobile discovery loses interrupt/wake registrars, task storage/session-close/account snapshot hooks, permission tables, copy/preference/model-discovery/composer examples, launch placement examples, VM projections and extra mobile-history scenarios. Retained checks cover abandoned versus dispatched fork uncertainty; /fast dispatch supersession; discovery lock release after exit; stale-send draft preservation; OSC status lifetime; future-version file preservation; launch cancellation before/during dispatch on both layout owners; mobile resume addressability before hooks; and recipe snapshot cleanup after plugin removal. The paired browser-link Electron source is unchanged. Terminal launch seams, mobile roles-only history and injected provider results do not establish native terminal destruction, transcript text fidelity or real remote provider cleanup. Audit3437 composes ROOT-adopted3431/independent3432: intentionally removes127 authored/174 potential expanded cases across12 sources, deleting7 whole suites and one267line unused helper without test credit. Retain8 authored/13 potential cases. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses reduced or deleted relay, coordinator mail, recovery watchdog, automation observer, launch picker, folder PTY, profile restart, web migration, retirement, pointer delivery, agent-row and run-bound mailbox examples. Retained checks cover relay hello/payload and SIGTERM drain, coordinator pending-check and clear-to-Run continuity, profile persistence, the full six-row web merge table and real run-bound delivery waiter. These do not replace removed alarms. Audit3429 composes root-adopted3426/independent3427: intentionally removes168 authored/205 potential expanded cases across16 semantic sources, deleting11 whole suites and one304line unused assertion helper. Five authored/five potential cases remain; the duplicate500k graph row is dropped while its50k row and callback remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted migration/recovery/federation/rich-draft/paste/automation/mobile/catalog examples. Keep server-edit preservation during migration, native supervised cancellation holding the agent home until physical stop, pagehide lost-text preservation, graph old-result WRAPPER WeakRef collection with50k rows, and mobile React render isolation. The WeakRef does not prove all commit objects are collectible or a memory budget. Audit3425 composes root-adopted3420/independent3421: intentionally removes83 authored/112 potential expanded registrations across13 semantic sources, deleting7 whole conventional suites and1 registering helper;7 authored/10 potential expanded cases remain. No equivalent replacement or fresh execution is claimed. Broad unit discovery loses the adopted mock orchestration, refusal, restart and storage examples. Retain unchanged two-phase real-WSL deployment/survival, the reduced idle-release holder proof without restart/first-output measurement, real worker stop/release cleanup barrier rows, physical Git branch-preservation rows, and SSH disconnected-close proof. Other SSH and CLI variants and WSL telemetry/warmup churn are intentionally removed. Audit3417 composes root-adopted3415v2/independent3416v2: intentionally removes117authored/173potential expanded registrations from12semantic sources, deleting4whole conventional suites and3registration modules and retaining4authored/5expanded cases, without equivalent replacement or fresh execution. Broad root-unit discovery loses naming/title/queued-host fixtures, main-window focus-shortcut and attach-services mock orchestration, Git Bash console-capacity fake process inventories, most real-Git preparation/locking variants and Claude/Codex catch-up examples. Keep lifetime rejected-writer followed by late termination proof and autonomous real-checkout cleanup; two-row real-Git rollback after a real move preserving a replacement/admin marker or revoked permission; a live translator write during a held transcript read fencing stale Claude history append; and Codex expected-item revision rejection before a historical writer shim. Codex driver registers retained bounds exactly once with original inputs; bounds does not prove modern writer persistence. Accepted changed-lock preflight sentinel and omitted-parent routing losses are not replaced. Audit3413 composes root-adopted3405/independent3406 and3409/independent3410: intentionally removes337 authored registrations/363potential expanded cases from13sources, retaining31authored/32potential cases and deleting one whole clipboard suite, without equivalent replacement or fresh execution. Broad root-unit discovery loses35WSL examples,22STT examples,22shared-server-banner examples,18clipboard-handler examples (whole suite deleted),14VM lifecycle examples and16commit-message cancellation examples. Retain4WSL and7STT mocked connector/worker cases,1rendered stale-banner refusal,2VM guard/failed-compensation cases and1late-stream ownership cancellation case; real WSL/microphone/remote-provider proof is not established. Separately discovered e2e-unit drivers lose13cross-version-wire local cases and15mobile active-publication cases, keeping same-process disk-reopen/new-host cursor content/epoch/sequence outcomes and a host/renderer-snapshot selection and inventory revision race respectively. Imported released stop-note scenarios remain unchanged and receive no duplicate credit. Cloud suites are outside root UNIT_INCLUDE. Audit3403 composes root-adopted3393/3394,3400/3401 and3398/3399: intentionally retires167 authored registrations/196 potential expanded cases across93 whole suites without equivalent replacement or fresh execution. Retained65 registrations/79 potential cases include one conditional missing-input failure sentinel; potential counts are not executions. Historical notes and complete dated receipts retain their exact original scopes. The broad root-unit command loses the UNIT_INCLUDE src/ subset of retired examples; mobile is outside UNIT_INCLUDE. Other lost alarms are recorded by the adopted per-case source reviews. Audit3390 composes root-adopted3382/independent3383 and3386/independent3387 and3388/independent3389: intentionally retires216 authored registrations/230 expanded cases in42 whole authored suites plus5 zero-case registration drivers, without equivalent replacement. Lost local alarms include fake PTY identity/routing/startup/replay/mode/title/mouse/IME/scroll matrices, broad startup identity and file-open wiring, mock mailbox title recovery, mobile row parity and whole-v-selective save parity. Retained24 registrations/26 expanded cases in22 suites preserve actual xterm Windows buffer/wrap/scroll, typed TOML scalar, real Promise cancellation/settlement, HTTP Codex scanner/projector semantics, durable recovery admission/newer-schema bytes, federation completion/ACK/SQLite atomicity and failed stale-run-write versus imported durable history. Additional retained clock-domain scheduling, durable freeze/ACK/checkpoint ordering, ABA dirty intent, SQL atomicity, folder-domain retier/last-note deletion, scrollback durable no-sync staging, PanelBridge subscriber lifetime and reserved-send publication preserve narrower concrete behaviors. Current finite owner drift assessed in bounded defining bodies; retained callback bytes unchanged. Historical notes and complete dated receipts keep their original exact scopes. No new execution, equivalent replacement, native focus/device/pixel, live remote or global mixed-version guarantee is claimed. Audit3381 composes root-adopted3376/independent3377 and3379/independent3380: intentionally retires436 authored registrations/670 expanded cases, including79 whole suites and12877 test lines, without equivalent replacement. Lost alarms include scheduling, account/host/status/filter/format inventories, microphone/image cleanup, title/bell/capability lifecycle routing, access/impersonation/outcome matrices, paste/input resend/ledger/retention, stale completions/reload containment, trust-read failure and popup OAuth/security/geometry vectors. Retained30 registrations/34expanded cases in22 suites preserve narrower SQLite TEMP/TZ/WAL, HTTP transcript root resume, mobile original-index Choose/Submit, child environment/Git lock expiry, stale ACK/PTY lifetime/peer fairness/durable replay/output ordering, actual DOM paste/editor focus, DEC2026 conservation, filesystem rollback, owned buffer lifetime, actual OSC parser/tracker handback, POSIX Python UTF8/fallback, durable trust concurrency/version, OpenSSH verdicts and hidden Chromium POST/route lifecycle. Mocked boundaries limit claims. Current bounded defining-owner reads preserve the retained oracles; captured identities are not full transitive semantic review. Both component blind pins authenticated before exchange; final reconciliation and explicit postpin3380reason correction supplement control, without retroactive blind claims. Root unit command excludes mobile and separately admitted Electron probes; use existing respective runners. No fresh execution, native focus/device/pixel UI, live SSH/remote, equivalent replacement or broad mixed-version/cross-platform validation is claimed. Prior dated receipts retain exact original scopes. Audit3375 composes root-adopted3370/independent3371 and3373/independent3374: intentionally retires79 authored registrations/84 expanded cases, including39 whole suites, without equivalent replacement. Lost alarms include federation field/parser inventories; Claude rejected-key precedence, credential clone redaction, managed-versus-user hook preservation and CLI defaults; setting/status/activity projections, mocked SSH/notification/reveal/Linear refusal, fake-clock reconnect and noise; image-fold artifact visibility, paste eligibility, setup trust, scripted ACP error-to-journal publication and ligature cache tuple aliasing; plus supplied folder-runner uniqueness, mobile host-overlay/credential/update/preference races, status/markup/title/keyed-row projections, subscriber cleanup, cancellation/close/refusal/link-issue and mirror suppression alarms. Retained3373 merger contiguity, dual-clock overdue-resume exactly-once, rejected-paste IPC sink and annotation schema roundtrip remain bounded. Retained actual HTTP/OSC status composition, destructive sibling-filesystem isolation, external contrast invariant and fixture CSS pixel intervention remain narrower. Original reviewed87authored/92expanded become9/9; current OpenCode suite adds two preserved upstream cases and the synthetic-title suite gained one Cursor policy mock-call case, independently reviewed for removal in root3378; current90/95 become11/11. That Cursor suppression/ready-frame regression alarm is intentionally lost; it does not observe native title ownership or tab flicker. For3370 only the peer was independently blind, with author exposure explicit;3373/3374 both blind pins were verified. This broad root unit command selects the reduced root unit suite; UNIT_INCLUDE excludes mobile tests, which require the existing mobile runner; retained Playwright component-pixel spec uses its separate existing headless fixture runner and is not validated by the broad unit command. No fresh execution, actual agent/shell/provider, OS material/focus, historical import equivalence, mobile, mixed-version or cross-platform evidence is claimed. All dated receipts retain their original scope. Audit3368 composes root-adopted3363/3364 and3367/3369: intentionally retires181 authored registrations/196 expanded cases, including28 whole suites, without equivalent replacement. Lost alarms include hibernation attention/liveness/backoff/sequential teardown and inventory, host launch compatibility/legacy/Kimi/resource cleanup, terminal recovery parser/pause/quarantine/remount, window-visibility omission-fence lifetime, launch-option rendered Start, sleeping-resume saved recipe/WSL quoting/cancellation, plus removed project/window/input variants. Additional tiny-suite losses include sentence/label/field/default inventories, credential omission, Windows argument/environment mapping, URI decoding, diagnostics, provider classification and fake-clock callback/parser examples. Retained ten registrations/twelve expanded cases include two literal Unicode Mn/Cf zero-width range guards in the actual character-range algorithm and protect narrower delayed editor bytes/destructive ownership, mounted hidden collision repair, ten transport epochs, duplicate-command refusal, caller timeout typing, executor unknown-opener refusal and failed-chunk user bytes. This broad unit command selects the reduced suite; no fresh runtime, real provider/network/PTY, filesystem restart, native UI, historical import equivalence, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3362 composes root-adopted3358/independent3360: intentionally retires129 authored registrations/140 expanded cases, including four whole suites, without equivalent replacement. Lost alarms include supplied orphan-adoption retry/malformed inventory/cancellation; fake-terminal replay parser/wedge/disposal callbacks; all-host catalog metadata and Windows preferences; host keepalive/protocol/stale/settings failures; and removed exact-stop/suppression, source save lifetime, diff-note queue, rename/editor close, input flood/chunk refusal/limits, purge and creation examples. Retained18 registrations/21 expanded cases protect narrower ownership, generation, dirty-state, queue convergence, encoding and attachment-lease outcomes. No replacement or keeper quota claim. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY/xterm parser, disk restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3359 composes root-adopted source3355/independent3356: intentionally retires130 authored registrations/139 expanded cases, including eight whole suites, without equivalent replacement. Lost alarms include strict create naming/capability/legacy payloads; supplied transport host replacement, deadline reattach, create-handoff and split activation; agent-session fork/clipboard/platform flags; shutdown cleared-scrollback/unmounted bindings; simulator host-collision/selection/shutdown; and removed editor, catalog, mutation, launch and status examples. Retained12 adverse-ordering/state-consequence cases are narrower and do not replace all removed alarms. Two queue absences receive zero removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH/PTY, filesystem restart, native UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3349 composes adopted batches3343,3345 and refreshed3354: intentionally retires234 authored registrations/294 expanded cases, including13 whole suites, without equivalent replacement. Lost alarms include supplied terminal reconnect/replacement/typeahead/mirror/result and snapshot examples; synthetic render-desync persistence/recovery; worktree listing/owner/catalog/lineage/removal and activation/creation examples; host-scoped PR and task-record projections; editor diff/status/comment and UI send-target examples; automation observation/wake, supplied sleep/deadline clock and native-chat completion/notification/relaunch samples; held-tab layout and cleanup/progress guard examples. Retained adverse-ordering and wire-boundary cases are narrower and do not replace all removed alarms. The already-upstream visibility-resume deletion receives zero new removal credit. This broad unit command selects the reduced suite; no fresh execution, historical import equivalence, real provider/network/SSH, native process/power/GPU, rendered UI, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain their original scope. Audit3342 composes adopted source batches3333/3334,3336/3337 and3340/3341: intentionally retires180 authored registrations/263 expanded cases, including seven whole suites, without equivalent replacement. Lost alarms include corrupt editor hydration; mobile editor/diff publication and privacy projections; stale-resubscribe routing; model-picker projection/refetch/retry/fallback; editor check-details and frame property inventories; latched-pane cleanup/quiescence; client create replay identity/options/compatibility; and removed owner, async routing, terminal and layout examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts keep their original scope. Audit3335 composes adopted source unions3327/3328 and3330/3332: intentionally retires100 authored registrations/123 expanded cases, including ten whole suites, without equivalent replacement. Lost alarms include seeded host persistence/contention and stale-host cleanup; browser snapshot assembly; mocked wake/DPR callbacks and frame budgets; preflight projections; palette authority/cache suppression, URL intent and saved-title cross-repo collisions; metadata coverage/set patch and dropdown assembly; and removed terminal, worktree, review-refresh and owner-routing examples. Retained definitions do not replace these alarms. This broad root unit command selects the reduced suite; no fresh execution, historical source/import equivalence, real provider/network/SSH, rendered UI/GPU/focus, durability restart, mobile, mixed-version or cross-platform validation is claimed. Unrelated dated receipts retain only their original scope. Audit 3318 composes adopted sources 3310, 3313 and 3317: intentionally retires 71 authored declarations and 102 expanded cases, including whole terminal-detach agent-retention, seeded batch/single open-files equivalence, shallow context-menu shortcut-badge, mocked active-workspace delete-dispatch, workspace session hydration, SSH target cleanup, editor session hydration and worktree terminal purge suites. Lost alarms include chained transfer alias and immediate close/exit retention; batch/single divergence, survivor/selection/map preservation and input nonmutation across 400 imperative seeds (one registered case); disconnected pending-handle retry, authoritative absences/close-record shortcut/older-host absence and published-presence/unpublished-frame examples; published empty host-graph omission to pane exit; legacy search ranking/exclusion fallback; provider request upgrade/dedupe/finalizer ownership; asynchronous PR suppression/primary-change and same-number different-repository URL rejection. Deletion routing loses host/instance qualification, paired and legacy mapping, folder ownership, active-versus-hover/menu-row selection, ambiguity, modal/editable/creation/view/main/deleting/missing guards and shortcut advertisement. Unchanged real keyboard/hover/confirmation/catalog E2E remains narrower, not an equivalent replacement. Additional losses include scripted old-host disk/UTF-8/noop/baseline and landed/unlanded lost-reply/replacement-retry outcomes; ordinary/repeated healthy-link reattach; cross-owner terminal/chat binding; oversized replay permanent-unavailable classification; floating reveal keyboard ownership/close/disable/listener disposal; naming/retained-turn/Sleep/error/close/reject status tables; SSH membership prune/new-target-only reads and environment/local bucket isolation; same-frame receipt capacity and positive bare/headless publisher readmission. These actual regression alarms are deliberately lost; no fresh gate, provider, rendered, mobile, mixed-version, cross-platform, soak or measured-budget validation is claimed. Audit 3312 composes adopted sources 3302 and 3308: deliberately retires 33 authored declarations and 36 expanded cases, including the whole backend startup-focus, active-browser IPC-close-routing, remote-browser-creation and agent-terminal-presence suites. Intentionally lost alarms cover distinct six-pane reconnect handles; SSH host label/Connect callback, unhydrated visit cleanup, repo cache aliases and hidden worktree/tab/PTY cleanup; adopted-agent startup focus and viewer navigation/reuse; staged, pageless, source, stale and orphan browser-close routing; web local-materialization refusal, desktop incapable-host fallback, failed/thrown remote creation and paired-web sign-in refusal; BOM/NUL and remote/local custom-editor byte-read examples; immediate folder activation and unread clearing; and batching idle/coalescing, accumulated byte ceiling, leading-edge-disabled, standalone synchronous oversize rejection, validation yield, deferred pending-tail order and both viewport-clear examples. The newly reintroduced agent-terminal-presence suite also loses actual terminal-status membership positive and nonterminal-status callback suppression alarms, including the recent 40916 guard regression; the older different 2133 presence-suite history remains historical and receives zero duplicate removal credit. These real regression alarms are intentionally lost. The unchanged survivors provide narrower coverage, with no equivalent replacement or fresh gate, provider, rendered, remote/mobile, mixed-version, cross-platform, soak or measured-budget validation claimed. Audit 3301 combines adopted source batches 3296 and 3299: retire 27 authored declarations and 36 expanded scenarios, including six whole suites. Intentionally lost alarms cover failed-verification/recovery/restart counters and mocked hosting/toast; early overlapping pane-read scheduling/count; unreadable removal listing draft guard; bulk-purge remote-qualified and raw-local guard maps; initial-snapshot blank/mode/oversize/printable/fresh-spawn callback and replay examples; fallback mobile file close and ordinary MRU; default queue overload/drain; seeded browser-page reuse and supplied missing-page URL recreation; ambiguous sleep/hibernation owner refusal and untouched bindings/origins; ordinary selection/activity/fallback priority; skip-remote success and retained hints; supplied structured-chat unsent retention and scoped removal draft eviction; caller-chat focus precedence; open/draft PR cache eviction and merged-head fallback result inventories. Retained adverse-ordering and ownership cases are narrower and do not establish equivalent replacements. No fresh runtime, native, physical provider, cross-platform, mobile or mixed-version evidence is claimed. Audit 3292 combines adopted 3285/3287/3289/3293 removals: 40 authored declarations and 47 expanded scenarios, including whole frame-editor-view, ambiguous remote-delete-refusal, unverifiable-host subscription and editor-close cleanup suites. The previously reviewed dock-badge suite was already removed upstream in PR 40895 and receives zero new removal credit. Intentionally retired alarms include background browser selection and last-tab landing; main startup fallback, SSH project cleanup/survivor routing, exact-PR contradiction/saved-push positives, replay/sidecar rollback and wrapped remote error normalization; records/snapshot missing-diff hydration, ambiguous folder-owner delete refusal, focus no-op notifications, nested editor context-only cache check, pending broadcast-first status and ordinary four-confirmation aggregation; preview-source frontmatter pruning, approved local/runtime agent dispatch, non-GitHub legacy PR cache invalidation, stale merged-head transient-error fallback, empty-queue immediate query reply and unverified-to-concrete preflight risk. Additional intentionally retired alarms are injected host subscription/unsubscribe counts, synthetic editor reveal/draft/fallback/scoped close, ambiguous unread refusals and hidden mark/read roundtrip, immediate connect backoff/noError, positive mirrored evidence-clock renewal, supplied VM failure/success projections and git-local/folder-SSH wake permutations. Remaining temporal/ownership cases are narrower; no equivalent replacement or fresh execution is claimed. Audit 3283 removes 20 authored declarations and 31 expanded scenarios, including two whole suites. Retired alarms cover PR refresh-state pruning, shared runtime subscription cleanup and store isolation, language content-change routing and older completion ownership, ordinary panel retry-timer cleanup and pending loading, four hosted-review lifetime permutations, mocked terminal creation error/order wiring, five sleep-inventory variants with pending-frame-only assertions, palette fallback/composite highlights, Windows/UNC encoded-editor input rows, mixed-repository partial failure counts, title setters and generated-title bypass/replacement, five window-close coordinator delegation/guard/finally branches, and ambiguous-old-record frame reveal. These regression checks are intentionally lost under the test-value audit; retained cases are not equivalent replacements. Prior coverage notes and receipts remain historical. Audit 3273 removes 14 authored declarations and 19 expanded scenarios across four whole suites and eight partial suites. Retired alarms cover saved-local cancellation/layout ownership, ambiguous-owner warning reset, folder attachment routing, VM false-capability bypass, abandoned reveal warmup cleanup and viewport claim deduplication/rearming, quiet unrelated remote-removal refusal, startup-command one-shot consumption, repository/floating recovery-ledger rows, fake-React pinned-scroll intent, split-ratio state identity/subscriber suppression and fractional change, cross-host project-group join refusal, three PR request-lifetime permutations, upstream-status preservation on error, and unresolved Git-owner refusal. These regression checks are intentionally lost under the test-value audit; retained scenarios are not equivalent replacements. Historical coverage receipts remain historical, including gate35 subscriber claims. Audit 3256 removes these specific alarms from the broad root unit commands: skill dialog active-share selection, close and pending-share handoff; simple keyed-record deletion; marker live-region insertion, simple landing and nine baseline policy examples; General search/control navigation inventory; orphan-count cache bootstrap; seven Codex code classifications and HTTP 429 warning mapping; two skill unread-root verdict examples; peer-mail payload and conversation-ID rejection; and the GitHub organization TABLE text-field optimistic rollback/failed-envelope example. These include real behaviors; retained tests do not provide equivalent coverage for every removed scenario. Mobile losses are recorded separately and are not coverage of these root commands. Audit3246 retires NewRowsMarker reveal/no-repeat and pane-scoped beforeinput checks; skill-owner option availability/count/dedup/order; six styled Input/Textarea IME forwarding rows; the static Codex/OpenCode smart-key inventory; tab-map AST ratchet diagnostic fixtures; child-roster work-budget and retained-aggregate alarms; malformed persisted floating-cwd fallback/no-crash/empty exact-root checks; the unanswered-host empty-list discriminator; the direct invalidated-cache commit-false assertion; one redundant digit-keyup IME row; cached non-supported Claude title-generator rows; and the duplicate late naming-result singleton. These root broad commands lose those specific alarms without equivalent replacement. Distinct retained owner/state checks have their own bounded claims; the real scheduler still rejects stale response data, while the direct false-return alarm is retired. Audit3231 retires supported-host restore-pane key mapping, closed/open creation-sidebar toggle visibility, unrelated scope-error/null classification and quick-submit reveal/no-create checks. These broad commands lose those real narrow alarms; no replacement coverage is claimed. Audit3219 removes the synthetic Cmd+J extra-source suite; this broad command no longer checks its match/tie placement and selection, announced count, empty-query cap, or workspace hiding with retained recent tabs. No replacement coverage is claimed. Audit3209 composes approved sources3202/3206: deliberately retires seven authored/expanded cases in four whole suites without equivalent replacement. Work-item queries lose real GitHub/GitLab oversized ASCII secret-prefix empty-query and whitespace-before-trim refusal alarms; inputs share the imported size bound and exercise only its fast length branch, with no multibyte or actual provider/exposure guarantee. Skill sharing loses complete/partial managed-bundle and destination-ambiguity association checks plus three negative origin/path/scheme admission examples; these do not establish uploaded contents, arbitrary-URL fetching, authorization or execution. Retain exact originless-project direct-link fallback, cloud-delete confirmation, bundled-target repair/owner isolation and explicit Windows-host preference over stale WSL suites. Rejected3201 runtime-preference removal is excluded. No real cloud/network/host/disk, mobile or mixed-version execution validation is claimed. Audit3200 composes approved source3194: deliberately retires thirteen authored registrations/twenty-seven expanded cases in six whole automation host/list policy suites without equivalent replacement. Real RPC category/retryability and wrapped-timeout recognition, unscoped-failure/scoped-success health evidence precedence, exact-ID create selectors and per-run workspace omission, paired-web local/SSH/absent host catalog routing, incompatible/removed/unhydrated/offline execution health, runtime-before-nested-SSH recovery dispatch, and no-owned-focus/successor/foreign-list focus policy alarms are consciously lost. Retain exact20line orphan allow/refuse/precondition and exact90line cross-host same-ID search suites. No retry-timer/fence lifecycle, actual reconnect/network/write delivery, native focus, mobile/mixed-version runtime or disk persistence coverage is claimed. Audit3191 combines approved source3182/3187/3188: deliberately retires thirteen authored Vitest registrations and fifteen expanded cases, including five whole suites, without equivalent replacement. Attachment intake loses real portal/native-capture terminal interception and unowned-container refusal, immediate child-only committed availability/claim-ref negotiation and delivery, and direct-target over stable focused/background composer routing through disable/re-enable/remount and independent question focus. Exact destination/cache/draft and final-stat owner/workspace/File-release suites remain; they do not replace Surface recipient routing. Automation authoring loses destination runtime catalog membership, local/direct-SSH exclusion of paired-runtime repos, unknown-runtime empty catalog, logical project/host/setup/repo/path context payload, and missing/nonready setup refusal; these were genuine direct helper alarms, while exact schedule transformation, legacy schedule and setup-decision suites remain. Agent picker loses setup-chat/terminal/missing-installer/unsupported-or-unknown-host row projection samples and the ordinary unsearched menu variant; the entire original26-line resume recipe suite remains, including reported bypass versus default/absent permission modes, as does the exact search-result menu callback with its fixtures, model/effort commits and staged Arguments reopen outcome. Rejected predecessor3180 full recipe deletion is excluded. No actual native drag/cursor/focus/IME, upload bytes or disk/restart persistence, full automation save/scheduler/runtime transport, real resume-process flags/environment delivery, or native menu/rendered layout validation is claimed. Audit3178 combines source3169/3170/3175: deliberately retires eight authored units (six Vitest cases and two imperative custom relay scenarios), including two whole suites, without equivalent replacement. Voice Section loses per-model pending-delete isolation across interleaved A/B completion and the two-success refresh-call alarm; exact VoicePane delayed key-clear/newer enabled-toggle merge remains and does not replace row deletion coverage. Speech loses dedicated rating-scale/inclusive-ceiling, unsupported-versus-unmeasured, explicit/system-language precedence/locale alias, accuracy versus fast named/default/best-measured selection, and unmeasured/no-match examples; exact offered-user-pick versus retired-selection case and its fixture remain. Relay loses legacy ping-delivery count/liveness sampled before the silence deadline and grouped malformed Bearer/host-ID 401/429 routing checks; twelve real Worker/DurableObject/shipped-client scenarios remain, not equivalent replacement for those narrow alarms. Six prior speech cases are not all deleted: approved successor3175 keeps the original user-choice case. No actual credential/file persistence, native/GPU/engine/download delivery, production assignment policy or deployed-runtime validation is claimed by this reduction. Audit3165 combines source3160/3162: deliberately retires six authored/expanded cases across three whole suites without equivalent replacement. Launch/status losses: observed React-render pending/resuming coordination across settings and deferred offer read; stale-offer reread suppressing dialog reopen; held-local-chat recovery visibility with new-chat preferences off; disabled-local admission avoiding offer RPC despite an injected but unconsumed paired-tab fixture; and run-only lost-reply uncertainty reachability until direct viewed-run retirement. Launch probe does not prove arbitrary synchronous-store atomicity, the paired fixture is not remote runtime enrollment, and the status case directly marks viewed state rather than rendering the summary. Crew-hook losses: old-host method_not_found cache persists across party changes but is released by independently supplied connection generation/verified pairing status or pairing revision, with actual frame publication and teardown. These are real narrow coordination, admission, uncertainty and mixed-version cache alarms consciously lost. Exact web verified-runtime writer handoff guard remains; it does not replace those alarms or prove live host/native/mobile delivery. Audit3156 combines3147/3150/3152/3153/3154: deliberately retires39authored/40expanded native-chat recovery cases, eight whole suites and two orphan test-support files without equivalent replacement. Lost alarms: lost-reply unconfirmed reconciliation, skipped rows/counts and missed-frame repair, unknown/pending outcome and omitted-failure-list compatibility, uncertain history after dismissal, retry/refusal/dismiss lifecycle, hidden/consumed picker ownership, result-tree keyboard focus/disclosure and all-skipped retirement; unconfirmed-over-conflict guidance and failed-chat changes despite older host timestamps; nearest-listed ancestor/cyclic forest shape, descendant membership, host ordering/missing-host local fallback; skipped/readiness/admission counts, provisional skip/progress monotonicity, early per-chat verdicts with batch pending protection, older-host waiting, start-slot/delivery distinction, retry-over-failure and unknown-after-withdrawal attention; remote snooze/reconnect/reoffer, unseen mixed summary through open/close, watched completion retirement, new success beside old failure, failed-reread mixed-result preservation, filter/collapse completion resets, retry summary replacement and current-offer/history selection; cross-host workspace identity, rendered cyclic reachability, nested tri-state/toggle propagation, project eligibility isolation and collapse preserving selection. Exact fallback-list completion barrier, batched verdict/cleanup retention, stale-read versus dismissal publication race, and requested-but-never-rendered finished-run retirement remain; they do not replace retired alarms. Audit3146 combines source reviews3132/3133/3135/3138/3143: intentionally retire20authored/31expanded cases without equivalent replacement. Provider-session losses: structured-native terminal-resume exclusion, metadata-less completed identity, new-turn completion clearing, Pi provider-only/periodic/manual promotion, changed-transcript launch-config isolation, Pi/omp/prime-agent completed-session remote/manual-shutdown promotion, and quit/done precedence over heartbeat. Side-thread/question losses: finished-pane acquisition, hidden title/recap states, local/relay/relay-done conversation takeover with prompt/session/child resets, plain-message abandonment, compaction/session replacement and completion resets. Codex startup losses: incoming-owner cached child replacement, old-relay foreign live-session correction, completed foreign-root turn-fact permutations. Transcript losses: real temporary-JSONL partial sibling/final-child completion, unlabelled child attribution, compact cached replay preservation and100/1000completed-child roster retention/release. Compiler-ratchet losses: valid-line uncompiled-list parser grammar examples and injected-existence stale-path filtering; these did not exercise actual filesystem canonicalization or build exclusion. These are real lifecycle, ownership, replay, liveness and retention alarms, including recent#40246 regressions, deliberately lost. Retained conditional record replacement, clockless root/SSH owner isolation, persisted main-thread fork rejection, partial/final question-answer lifetime, cached-roster/missing-ID/live HTTP/durable startup boundaries and restarted-relay refusal/acceptance do not replace those losses. No native/SSH/WSL execution or equivalent coverage is claimed. Audit3128 scoped coverage reduction: Audit3121 removes VoiceDictationLanguageSetting.test.tsx (Spanish-to-es callback and unsupported-model Korean option presence) and voice-speech-model-rows.test.ts (provider/language-order/retired-on-disk list policy); these real narrow alarms are deliberately retired without replacement. VoiceSpeechModelSection per-model concurrent deletion remains unchanged. Audit3122 removes the direct absolute/parent-segment sparse-directory exception-type sample and the five-row IPC sparse-preset error-code/non-save table (two authored/six expanded cases). Those real typed-validation and IPC refusal/error-code alarms are deliberately retired. Keep both real symlink/junction floating-directory trust cases and the minimal RPC structured validation-reason serialization case. The RPC case does not replace IPC save/refusal behavior, typed normalization for both path inputs, actual Git sparse checkout or rendered localization. No equivalent replacement or fresh test/runtime/platform/remote/performance validation is claimed. Audit3124 removes ExperimentalPane.test.tsx (ordinary Pet switch callback), WorkspaceDirectorySetting.test.tsx (padded absolute saved-path normalization/no-write blur), and repository-search.perf.test.ts (navigation/selected-pane builder-call policy, fixture cardinality/nonempty entries and renamed project keyword inclusion). These real narrow alarms are retired without replacement; RepositoryPane.search.test.tsx search-to-sparse-editor request remains unchanged. Audit3129: resolver Windows case-insensitive and POSIX case-sensitive configured-env composition plus mocked-base launch-preflight Windows composition are intentionally no longer checked by the two deleted suites. No equivalent coverage claimed. Keep runtime-structured-agent-host-env.test.ts unchanged: configured Windows Claude-home account identity and host-account precedence remain checked. Audit3114 incorporates source3101: retire three structured Claude child-environment wrapper cases covering inherited stamp isolation/default and explicit entrypoint, Windows SDK-key canonicalization/default/duplicate-key order, and configured lookup/final environment/config-dir conjunction. Keep direct auth/session sanitation and Windows-versus-Linux overlay checks, actual executable lookup from supplied inherited PATH, native Windows child environment delivery, and stale option-mutation fencing. These are not equivalent replacements for the retired wrapper alarms. No fresh test/runtime, real login-shell probe, Claude SDK startup, remote/mixed-version, UI or performance result is claimed. Audit3114 incorporates source3105: retire 17 authored/20 expanded daemon launch-note cases, including the entire two-case SessionLaunchNote suite. Deliberately lose split-marker carry across same-token re-arm, stale-token retirement refusal, foreground tri-state and serial settlement, omitted-note/no-incarnation mixed-version policy, sticky retirement, first-incarnation/replacement launch preservation and held-launch precedence. Keep only the unchanged unresolved foreground-read race against a later finished attach. Supplied notes/foreground and runtime calls do not establish actual daemon/main restart, SSH, native Windows observation, mobile UI, equivalent replacement or fresh runtime/performance validation. Audit3114 incorporates source3109: retire six authored/35 expanded pricing cases: 21 accepted model-name examples, 10 unsupported/future-model fail-closed examples, malformed TTL clamp, unknown-model estimate refusal, flat-model long-context disregard and oversized long-context clamp. Keep mixed five-minute/one-hour write allocation and short/all-long/mixed long-context allocation with original fixed oracles. Real parser/refusal/bounds alarms are knowingly lost without equivalent replacement; retained tests do not validate provider rates, actual billing or request threshold classification. Preserve the archived usage-scan-append source-freeze receipt without transferring historical execution/performance credit. No fresh runtime result is claimed. Audit3108 source-only reduction (3094/3095/3100): retire 27 authored/30 expanded cases, including complete markdown view-state, file-preview link-open and CLI agent-server-route suites, without replacement. Deliberate losses include preview visibility/reference lifetime and same-path owners; SSH stat, directory/missing-target, explicit-null owner, nonmarkdown line/column, local URI routing, remote-to-client guard and same-file reuse; document-preview source-owner/folder/reuse routing, stable grant location and outside-workspace POSIX/Windows request projection; and verified remote CLI send/check refusal, suggested --on argv/copy, shell no-probe bypass, unverified-launch admission and older-peer refusal. Retain two markdown source-runtime/qualified-reveal alarms, one stale grant-rejection fence and two remote read/mutation evidence policies, with original callbacks. Final last-wins reveal did not prove transient cancellation; CLI prerequisite did not prove no real send, and envelope tests do not execute server refusal. No fresh runtime, rendered UI, actual file/SSH/paired transport, authorization, mixed-version or performance result is claimed. Audit3096 source-only reduction (3087/3088): retire 69 authored/76 expanded cases, including complete recovery-portals and reconnect-banner suites. Intentionally lose dead/default TUI replay, generated alternate-frame style composition, reset/mode/dimension and legacy missing-mode-boundary alarms, UI notice/action/reveal policies, and isolated recovery lifecycle variants, without equivalent replacement. Retained registrations and callbacks are unchanged; headless cells are not pixels, and no fresh runtime, host, SSH, WSL, rendered-app or performance result is claimed. Audit3086 source-only reduction (3079/3080/3081): intentionally retire 93 authored/100 expanded mock-boundary cases, including the complete spawn-errors and SSH reconnect overlay suites, without replacement. Retired alarms include spawn classification/ownership, reply ceilings and renderer-write failure recovery, disconnect/detach fencing branches, authority facts, and overlay retry/visibility wiring. Retained callbacks are unchanged; no fresh execution, real PTY/SSH, rendered UI, or performance result is claimed. Audit2432: The broad unit command loses the isolated BranchPicker open-change notification sample. Its unique true/false callback alarm is intentionally retired; reopening/status-refresh behavior was not exercised. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2425: The broad unit command loses the workspace-copy-status supplied readError forwarding sample. It did not exercise the branch-switch blocking promised by its title. This alarm is intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2417: The broad unit command loses two ordinary clean Git cherry-pick/revert success samples and two terminal startup foreground-read work-budget samples. These unique alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2412: The broad unit command loses the whole six-case fake-React-hook scroll-anchor suite, two generic draft-retention samples, a simple new-agent order sample, a real history-publication batching-count sample, an unmounted handoff abort/timer sample and three basic WSL preview/error samples. These alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2404: The broad unit command loses the standalone share-name lookup 503 propagation suite, initial deletion placement-kind sample, direct changed-mtime stale sample, first-pass rollback existence example and abandoned-preparation directory sample. These exact alarms are intentionally retired without equivalent replacement or fresh validation of the reduced selector. Audit2397 review amendment: The broad unit command loses three standalone suites, two partial-case definitions and two URL table rows across package/download/placement tests. Keep initial wrong-port authorization, modified independent-copy digest refusal and abort-after-begin cancellation dispatch. The remaining alarms are deliberately retired; no equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2393 current-parent refresh2401: Remove the basic skill-install service alias sample and original public managed-remove receiptless refusal case. Preserve the upstream request ownership suite and transaction interruption coverage. No equivalent complete replacement or fresh validation of the reduced selector is claimed. Audit2389: The broad unit command loses the missing-preview suite, plain-global destination sample and prepublished-receipt recovery example. No equivalent replacement or fresh validation of the reduced selector is claimed. Audit2385: The broad unit command loses four authored walker/staging/protocol examples (six expanded cases); no equivalent replacement or fresh validation is claimed for the reduced suite. Audit2383: The unscoped Vitest command no longer includes the removed Antigravity catalogue suite; its exact provider-root/owner example has no equivalent replacement or fresh validation claim. Audit2377/2381: The unscoped Vitest command includes the reduced bundled-skill and native-discovery suites; the eight removed backup/legacy/catalogue/symlink/SSH-builder examples have no replacement-equivalence or fresh validation claim here. Audit2133 current-parent preparation: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Audit2095 current-parent preparation2126: Current reduced sources/selectors intentionally retire the alarms listed in knownGaps; prior dated runs and replacement claims do not validate the changed set. Historical prior account (preserved verbatim): Require both existing named browser placement scenarios to pass three times with one attempt, zero skips, zero failures, and no report errors.", "commands": [ "gh workflow run packaged-browser-e2e.yml", "pnpm exec playwright test tests/e2e/packaged-mixed-version-browser-placement.spec.ts --config tests/playwright.config.ts --project=electron-headless --workers=1 --repeat-each=3 --retries=0", @@ -14062,6 +14066,7 @@ "Collect repeated scheduled history before making this required." ], "knownGaps": [ + "Audit3563 scope: retire the whole Kiro OSC/hook identity suite and the OpenCode2 identity-only case; reduce Codex cached startup child-retirement and missing-session-ID tables to one case each (2 authored/4 finite removals). Intentionally lose Kiro alias/session continuity, OpenCode2 completed OSC identity continuity, cached replacement without relay-generation restart, and missing-previous-ID context-preservation alarms. Remaining provider, remote compatibility, persistence and cross-session pollution cases are not equivalent replacements. Also remove the private Codex roster-size assertion while retaining public child-absence checks. Historical evidence retains its original source scope; no new execution or transferred pass validates these reductions.", "Launch direct-owner cleanup current scope: This gate’s unfiltered unit command now selects a suite reduced by 82 whole registrations / 159 potential table examples in 10 whole test Programs, including 8 registrations / 15 examples from five once-registered create-message factory closures. Lost alarms include frozen first-message/options and legacy-host fallback; distinct actions and redelivery; uncertain/refused/reloaded input custody, Retry and late proof; rich prompt/attachment upload Stop/close and handback; ordinary draft/editor/IME/StrictMode lifetime and accepted topology; provisional fallback tab placement/group/pin/order and rich worktree handoff. Four whole consent/resume-identity cases remain in two Programs; they do not replace these alarms. This entry governs current reduced-unit scope and qualifies any broader implication of unchanged oracle/coverageNotes history. Entire dated evidenceRuns, original commands/currentCommand values, counts/results and existing history retain their original scopes only; no fresh execution, transferred pass, replacement equivalence or prior upstream-removal credit is claimed.", "Audit2133 current-parent preparation (988 authored removals; independently reviewed2099/2107/2115/2116 decisions; upstream clear-buffer and recovery-image regressions retained): Intentionally retire these low-value alarms without replacement: src/main/persistence/worktree-sort-order-write.test.ts: re-ranks every writable row when a blocked row sits in the requested order. src/main/runtime/construction/orca-runtime.test.ts: remove only literal registration import/call pairs await import('../orca-runtime-tests/worktree-setup-and-startup.spec'), await import('../orca-runtime-tests/gitlab-and-pr-bases.spec'); no additional authored credit. src/main/runtime/orca-runtime-tests/targeting-and-resilience.spec.ts: routes tab switch and capture start by explicit page id; accepts focus on tab switch without altering bridge args (focus is main-side concern); keeps terminal cwd resolution fail-soft when the provider is unavailable. src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-04.spec.ts: drops the bounded scan cache when orphan-cleanup removal completes; clears optimistic reconcile tokens when a CLI worktree removal succeeds; emits drift without mutating when the fetched base fast-forwards created HEAD; emits base_changed without mutation when the fetched base rewrote history; skips stale-token reconciles without mutating or emitting stale status; emits unknown without mutation when fetch fails or the base ref is missing. src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-03.spec.ts: routes runtime preserved-branch force-delete through the selected WSL project runtime; replies to a waiting client with the error of a delete that fails after acceptance; treats forced runtime deletion of an already-missing unregistered worktree as cleanup; treats normal runtime deletion of an already-missing unregistered worktree as cleanup. src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-02.spec.ts: warns that a missing-repo removal only forgot the workspace; restores runtime watchers when CLI worktree deletion fails after teardown; does not re-spend the drain budget on a removal after a wedged install was abandoned; routes runtime worktree removal through the selected WSL project runtime; force-deletes a branch that was preserved by runtime worktree removal. src/main/runtime/orca-runtime-tests/gitlab-and-pr-bases-part-02.spec.ts: records GitLab pasted-project recents only after successful runtime lookup; threads explicit origin preference through runtime WSL PR base resolution; resolves local GitLab fork MR bases from the target project MR head ref; captures the fork MR head from a dedicated ref, not the shared FETCH_HEAD; keeps the durable MR head when the head fetch fails but the local ref resolves; routes runtime GitLab fork MR base git calls through the selected WSL project runtime. src/main/runtime/orca-runtime-tests/gitlab-and-pr-bases.spec.ts: passes SSH connection ids through GitLab task operations; routes runtime GitLab issue, MR, work-item, and todo actions through the selected WSL project runtime; routes runtime GitLab MR details, review-management, job, and pasted URL actions through the selected WSL project runtime; normalizes runtime GitLab issue list arguments like the desktop IPC path. src/main/runtime/orca-runtime-tests/worktree-setup-and-startup-part-03.spec.ts: does not warn when setup is explicitly skipped for CLI-created worktrees; puts the agent a create starts in the first default tab, as the window does; keeps the 8s main-runtime startup readiness budget for agents without an override; gives OpenCode the 20s main-runtime startup readiness budget; launches explicit startup agents with prompts for CLI-created worktrees. src/main/runtime/orca-runtime-tests/worktree-setup-and-startup.spec.ts: returns a setup launch payload for CLI-created worktrees when hooks are explicitly enabled; does not hand the window a setup main never tried to start; passes the selected Windows setup shell into runtime runner generation; follows normal setup policy for CLI-created worktrees without activating them; uses returned WSL setup shell metadata when runtime spawns setup; uses the shell-aware setup runner for windowless creates without a startup command; reports the in-process setup hook as running when nothing can launch the runner. src/main/runtime/orca-runtime-tests/terminal-sleep-and-teardown-part-02.spec.ts: coalesces two clients sleeping the same host worktree; stops exactly the expected live PTYs for a worktree; explains a refused workspace close instead of throwing the raw refusal enum. src/main/runtime/orca-runtime-tests/mobile-summaries-part-03.spec.ts: marks the desktop-active worktree as isActive; includes SSH-backed worktrees in the mobile worktree summary; caches repeated malformed path misses before normalizing them again. src/main/runtime/orca-runtime-tests/mobile-summaries-part-02.spec.ts: keeps title-only foreground work ahead of monitoring in another pane; keeps a local hook row while a connected PTY still backs its pane; keeps an OSC row via its connected PTY after the pane binding is cleared. src/main/runtime/orca-runtime-tests/mobile-summaries.spec.ts: resolves files through a source-visible linked-checkout worktree; marks saved session tabs with live PTYs as host sidebar activity; attributes live legacy PTYs from saved layout bindings when their panes are hidden; marks saved browser tabs as host sidebar activity like desktop; falls back to the path-keyed GitHub cache entry; includes folder workspaces in compact worktree summaries for mobile; attaches inline agent rows from the latest OSC 9999 status; projects monitoring over a title-derived working status. src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-03.spec.ts: pushes to an idle pane when the only live waiter filters out the message type; resolves a registered waiter without PTY-injecting when the leaf is already idle; keeps already-idle status after tui-idle wait for immediate message delivery; resolves message waiters when notifyMessageArrived is called; does not resolve type-filtered message waiters for unrelated message types; removes message waiter abort listeners after message arrival; resolves message waiters on timeout when no message arrives. src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-02.spec.ts: delivers pending mail via notifyMessageArrived when the recipient is already idle; does not keep retrying when every pending row was already pointed; repoints pending rows restored with a live-idle coordinator. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-13.spec.ts: rolls back a half-created terminal whose surface never publishes. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-12.spec.ts: rejects startup prompts for agents that require post-ready stdin; uses portable Unix quoting for mobile agent launch commands in WSL project runtimes; keeps PowerShell quoting for mobile agent launch commands in Windows host runtimes; uses cmd.exe quoting for mobile agent launch commands in local Windows host runtimes; publishes headless mobile session agent identity with synthesized PTY status; rejects disabled mobile session agent launches before spawning; validates mobile terminal insertion anchors before resolving agent launch commands. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-11.spec.ts: coalesces a reconnect close burst onto one authoritative PTY inventory; keeps an explicit user close destructive while the PTY is live; builds mobile session agent launch commands on the runtime host; injects mobile quick-command prompts into the host-built agent startup command. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-10.spec.ts: closes persisted headless terminal parents before any prior list call; tears down a serve-owned headless tab on close while a renderer is attached so it cannot resurrect; defers a renderer-published pending tab to the renderer instead of tearing it down; returns a delegated close outcome with no selection tombstone when the renderer owns the outcome; tears down a runtime pending shell the renderer never adopted on close. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-08.spec.ts: materializes a plain shell when the pending tab has no launch agent; hydrates legacy persisted terminal tabs without layout entries; reads and indexes the full agent-status snapshot once per mobile projection; kills persisted SSH PTYs when closing hydrated headless tabs before pane metadata is restored; materializes hydrated pending headless terminals with the persisted session identity. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-07.spec.ts: hydrates a persisted SSH-owned pane before an attached renderer publishes its graph. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-05.spec.ts: publishes the desktop's tab name over the live agent title, and the live title without one; uses fresh neutral PTY titles over stale mobile snapshot and OSC titles; operates PTY-backed mobile session terminals without a renderer graph; reuses pane close for live PTYs that do not own a renderer tab; lists PTY-backed mobile session terminals without a renderer graph; shows and resolves active PTY-backed mobile session terminals without a renderer graph. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-04.spec.ts: creates mobile session terminals for folder workspaces in a headless runtime server; spawns fresh headless SSH mobile session terminals instead of reattaching synthetic local ids; hydrates headless mobile session terminals from the host workspace session; keeps live headless mobile session terminals when a desktop renderer publishes without them; publishes laptop-created remote runtime split terminals to phone session tabs. src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-02.spec.ts: does not report headless wake degradation without sleeping records; does not wake slept agents for non-mobile session-only activation; does not rewrite unread metadata when a mobile activation finds the worktree already read; returns unread:false from worktree.ps after a mobile activation clears the flag; materializes pending mobile session terminals without focusing desktop clients; closes browser mobile session tabs when addressed by browser workspace id. src/main/runtime/orca-runtime-tests/mobile-session-tabs.spec.ts: resolves mobile terminal surfaces by exact split leaf; closes the matching mobile terminal UUID leaf without closing the whole tab; closes the whole mobile terminal tab when addressed by parent tab id; activates the active split leaf when addressed by parent tab id; clears unread metadata on mobile worktree activation without focusing desktop clients; wakes slept agents on the host renderer when a phone activates a worktree. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-05.spec.ts: publishes the hook provider session on a headless mobile tab so native chat can address the transcript; recovers the agent type from the hook row when the pane was launched without an agent hint; reads one agent-status snapshot per projection, not one per terminal tab; waits for unknown-launch foreground owner before publishing Pi-compatible mobile status; keeps Pi-compatible mobile session status as Pi for an unknown-launch foreground pi PTY; keeps renderer-vetted mobile agent status for custom-titled terminals. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-04.spec.ts: publishes terminal surface agent status for paired web clients; preserves authoritative OMP identity for Pi-compatible remote terminal snapshots; normalizes a remote OMP title without republishing omitted launch identity; skips the foreground-process probe when the PTY launch agent is already known; probes the foreground process only on a status transition for unknown launch agents; passes the client signal and the slept-worktree refusal to the spawn. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-03.spec.ts: does not use stale runtime-created PTY status when a neutral PTY title exists; recognizes runtime-created Codex PTY handles from the ready prompt; recognizes runtime-created Antigravity PTY handles from the ready prompt; recognizes live leaf Antigravity terminals from the ready prompt; recognizes a later Antigravity header with a prompt but no model line; keeps mobile terminal surfaces visible while their leaf handle is pending. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-02.spec.ts: lets Claude agents management titles clear stale runtime-created title status; does not recognize live Claude agents panes from a Claude foreground process; does not let stale tab-level agent titles override current neutral pane titles; lists live terminals with fresh pane titles over stale tab titles; lets adopted neutral pane titles use non-shell foreground fallback; lets adopted neutral pane titles retry wrapper foregrounds until recognized; authorizes settled CLI prompts only after positive %s foreground identity; does not recognize arbitrary foreground TUIs as running agents; lets live neutral pane titles retry wrapper foregrounds until recognized; lets current Claude agents PTY titles override stale runtime-created OSC titles; does not let stale Claude agents OSC titles suppress current PTY title activity. src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status.spec.ts: keeps preallocated terminal handles valid across renderer reloads; preserves runtime-created PTY process identity after graph unavailable; keeps prompt bindings fenced across runtime restarts without provider incarnation; recognizes runtime-created PTY handles with agent launch titles; does not treat a bare Cursor Agent native title as a running agent session; uses stale runtime-created PTY status when there is no title or foreground evidence. src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-07.spec.ts: keeps orphaned list and show writability aligned with the send gate. src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-04.spec.ts: cancels a coalesced SSH worker recovery retry when its provider disconnects. src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-03.spec.ts: retries renderer reveal before clearing an adopted legacy worker sleeping record. src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-02.spec.ts: recovers exported ORCA_TERMINAL_HANDLE from discovered live PTY sessions. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-12.spec.ts: applies ANSI erase-from-start line controls in retained previews; applies ANSI stripping for private or intermediate CSI line controls; applies ANSI stripping for unsupported erase-line modes; reveals an agent terminal under the pane the caller reserved; keeps a fresh %s pane host-owned when a desktop window is available; keeps desktop publication %s separate from the reserved host spawn; mints its own pane when the reserved leaf is malformed, and reports the one it used; starts an agent terminal with the model the caller picked; starts an agent terminal without a model flag when no pick was made; attributes a startup-agent launch named %s as %s; attributes a fresh agent session the host builds as unknown; leaves a bare agent command the user typed out of launch attribution. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-11.spec.ts: trims oversized terminal output bursts without per-line array shifts; builds terminal previews without mapping the full retained tail; keeps recent PTY replay output capped without needing previous data for large chunks; applies terminal redraw controls before retaining previews; applies ANSI terminal redraw controls before retaining previews; retains multi-line ANSI footer redraws instead of appending old frames; keeps the cursor column when erasing full lines in multi-line redraws; applies ANSI split redraw controls without leaking raw params. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-10.spec.ts: returns renderer visible screen lines through terminal.read RPC JSON result; separates composer drafts from provider-owned terminal screens. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-09.spec.ts: reports an in-flight PTY focus as not navigated when its notifier disappears; reports graph-leaf focus as not navigated without a host notifier; grounds the provider mode tracker before a later chunk, in the emulator order; grounds an in-flight provider snapshot capture so it cannot publish the pre-reset screen; waits for terminal exit and resolves with the exit status; keeps terminal read payloads bounded while retained output remains pageable; trims terminal read preview character budget without per-line array shifts. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-08.spec.ts: submits a silent Claude composer once after the bounded render fallback; chunks large terminal.send text before provider writes; yields chunked terminal input through immediates between writes; yields while validating accepted large terminal.send text before provider writes; reveals a background terminal session when focusing its handle; replays captured launch config when focusing a background agent session; reveals background terminal sessions with the freshest PTY title; renames background terminal handles without requiring a visible tab; keeps a background terminal handle stable while reveal adoption is racing. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-07.spec.ts: resolves tui-idle when a stale Codex prompt is followed by Antigravity readiness; returns an agent-neutral blocked wait result for update prompts; returns an agent-neutral blocked wait result for workspace trust prompts; returns an agent-neutral blocked wait result for cwd selection prompts; returns a blocked wait result for Codex model migration prompts; returns a blocked wait result for Codex startup hook review prompts; returns an agent-neutral blocked wait result for generic interactive prompts; splits text and enter writes for background terminal handles; types the lead line for %s: %s; submits through the agent-specific PTY timing policy for %s. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-06.spec.ts: waits for exit on background terminal handles; keeps retained PTY transcript memory when controller refresh fails; resolves tui-idle from an Antigravity ready prompt preview; resolves Antigravity ready prompts with newline-heavy pasted tails without splitting; resolves live-leaf tui-idle from an Antigravity ready prompt preview. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-03.spec.ts: spawns the available Qoder command for a captured same-session resume: %s; does not use the local Windows shell setting for remote Windows bare agent creates; matches canonical bare agent commands when a command override is configured; keeps non-bare agent command terminal creates unchanged; keeps disabled bare agent command terminal creates unchanged; injects runtime hook receiver env into terminal sessions; creates background terminal sessions for a $label; creates background terminal sessions for a $label; enables Claude Agent Teams only for direct Claude launches when configured in-process; reveals Claude Agent Teams launches with the rewritten launch config; preserves Claude Agent Teams for sequenced Claude launches. src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness.spec.ts: creates visible terminal sessions without asking the renderer to focus a tab; asks the pty controller for the requested shell instead of a startup command; knows the agent a flagged launch command names before any process read; retires a flagged command pane identity when the command exits; retires only receipted restored PTY authority on command completion and exit; forgets synthetic handles when disconnected PTY records are pruned. src/main/runtime/orca-runtime-tests/headless-snapshots.spec.ts: returns OSC titles from headless main terminal snapshots; adopts renderer-seeded titles into headless main terminal snapshots; returns cwd metadata from seeded headless main terminal snapshots; adopts OSC7 host metadata from seeded headless terminal scrollback; falls back to the renderer snapshot for hidden-output recovery without headless state; keeps an empty headless snapshot authoritative for hidden-output recovery; advances the absolute output sequence across a daemon stream gap; emits explicit OSC 9999 agent status from runtime PTY data. src/main/runtime/orca-runtime-tests/terminal-side-effect-facts-part-02.spec.ts: falls back to the provider snapshot for a restored PTY with no mounted renderer; falls back to the available mirror when authoritative provider history is unavailable; keeps an empty renderer snapshot when the provider has no retained content; keeps restored provider history authoritative after later live output; uses provider alternate-screen state while a partial model is unsafe; prefers the tracked title over the headless emulator lastTitle; keeps the cursor-agent literal in record-fallback snapshots only without a tracker title; touches mobile snapshots once for decorative spinner ticks, again on idle. src/main/runtime/orca-runtime-tests/local-worktree-creation-part-02.spec.ts: creates a same-repo PR branch override from a resolved head SHA and matching push target; skips broad remote fetch for an existing full-SHA PR base; creates a selected Bitbucket PR branch override from a matching remote branch; suffixes an existing PR when a matching push target lacks selected PR metadata; suffixes a matching push target branch when selected PR metadata has no PR number; suffixes a matching push target branch when the existing PR is different; suffixes a selected PR remote conflict when the PR lookup fails; checks out an unused runtime PR branch only when it is at the resolved head SHA; suffixes only the runtime worktree path when an exact PR branch checkout path exists. src/main/runtime/orca-runtime-tests/local-worktree-creation.spec.ts: returns runtime local base update suggestions from addWorktree; creates a runtime local worktree from a usable persisted local branch base; does not create a runtime local worktree when the refresh fails and no local base ref exists; creates a branchNameOverride worktree from the selected matching remote base ref. src/main/runtime/integrations/skill/orca-runtime-skill-recovery.test.ts: continues skill management after startup recovery fails. src/main/runtime/orchestration/mailbox/routing/orchestration-mailbox-detached-routing.spec.ts: uses bounded ownership lookup for direct arrivals while the agent cannot take input; wakes a bound Run waiter when same-Run direct mail arrives; keeps an unowned detached-recipient message in its direct mailbox; uses primary-key lookups for bounded routing updates. src/main/runtime/orchestration/mailbox/terminal-pointer/orchestration-mailbox-delivery.test.ts: remove only literal registration import/call pairs import './orchestration-mailbox-pointer-cli-command.spec'; no additional authored credit. src/main/runtime/orchestration/mailbox/terminal-pointer/orchestration-mailbox-pointer-cli-command.spec.ts: renders the %s CLI command in a mailbox pointer. src/main/runtime/rpc/methods/terminal/terminal-multiplex-link-estimate.test.ts: sizes a slow high-latency link near twice its bandwidth-delay product; never paces a LAN, even when its client parses slowly; never paces a fast link; returns to legacy windows when the path RTT jumps. src/main/runtime/rpc/methods/terminal/terminal-snapshot-scrollback-rows.test.ts: steps down the row ladder when history is over the byte budget; tries the larger first row count, then the usual ladder within the same byte budget. src/main/runtime/rpc/server/pairing/network/windows-default-route-interfaces.test.ts: returns unknown when route inspection fails or emits malformed data; does not inspect routes outside Windows. src/main/runtime/rpc/server/pairing/pairing-offer-owner-cli.test.ts: is refused to a paired %s client; reports a failed widening instead of advertising an unreachable address. src/main/runtime/rpc/terminal-multiplex-end-verdict.test.ts: reports exited only when the owning runtime completes the exit waiter; reports unverifiable when the runtime cannot observe the exit waiter; reports unverifiable when a disconnected PTY resolves with unknown liveness. src/main/runtime/session-graph/headless-runtime-graph.spec.ts: lets a session tab inventory settle on a host no renderer publishes for. src/main/runtime/session-graph/runtime-graph-standing-contracts.test.ts: remove only literal registration import/call pairs import { registerHeadlessRuntimeGraphTests } from './headless-runtime-graph.spec', registerHeadlessRuntimeGraphTests(); no additional authored credit. src/main/runtime/session-tabs/structured-conversation-tab-title.test.ts: bumps the stored snapshot version only when the title changes. src/main/runtime/session-tabs/terminal-close-observed-exit.test.ts: keeps the healthy aggregate verification and delayed physical exit behavior; uses a stamped exit for the incarnation that was actually being closed. src/main/runtime/terminal/identity/terminal-identity-probe.test.ts: answers live for a structured worker without touching the PTY graph; answers live for a PTY handle the runtime still holds; runs the full leaf check for a handle with no live PTY; answers not-live for a stale handle; propagates \"could not look\" rather than reporting it as a dead handle. src/main/runtime/terminal/layout/paired-desktop-server-writes-layout-owner.test.ts: names the tab in main’s layout through terminal.rename; takes a window-style layout.apply close as already stopped, so the terminal runs on. src/main/runtime/terminal/lifecycle/pty-lifecycle-generation-retention.test.ts: retains no lifecycle generation after a spawn/exit cycle. src/main/runtime/terminal/lifecycle/run-terminal-user-open.test.ts: reads a freshly spawned terminal as used once a user opens its tab, not on a probe; keeps the open mark on the process a user open starts. src/main/runtime/terminal/surface/terminal-surface-replies-window.test.ts: from a desktop window that mounts both panes; from a server with no window. src/main/runtime/terminal/surface/workspace-layout-projection.test.ts: carries views for a partition that holds none yet. src/main/runtime/terminal/view/terminal-subscribe-exit-waiter-leak.test.ts: releases a signalled exit-waiter when the signal aborts. src/main/runtime/terminal/wait/orca-runtime-exact-worker-provider-session.test.ts: selects a local hook status for a local pane. src/main/runtime/terminal/wait/prime-agent-screen-readiness-transcripts.test.ts: keeps the footer and the bare caret on screen mid-turn; %s: the screen alone reads ready during %s. src/main/runtime/workspaces/startup/runtime-worktree-agent-startup.test.ts: uses per-launch arguments and preserves launch telemetry; attributes a startup agent whose caller named no surface as unknown; launches %s bare for an absent or blank prompt; attributes a %s draft launch named %s as %s; merges an automation's extras over the host defaults; threads worktree-create extras into the startup build; returns an enabled requested agent without detecting; returns the default when nothing was requested; detects instead of using the default when the requested agent is disabled; starts no agent when the default is blank; detects an enabled agent on this host when nothing usable was named; starts no agent when detection finds none. src/main/runtime/workspaces/startup/runtime-worktree-startup-readiness.test.ts: pastes and submits through composer readiness, never typing raw input; types the follow-up on WSL or SSH-to-Windows once the agent is in front, as before; cleans up the deadline when the composer was already captured; a draft that is not submitted is pasted on the box cursor, as before; takes the box after the grace when the row never comes, whatever repaints; takes the box at the deadline when the grace would outlast the budget; lets a dialog on screen veto a grace; refuses the deadline inside a pending grace while a shell is in front; takes the box at the deadline inside a pending grace with the agent in front; takes the box at the deadline after a refused grace once the agent is in front. src/main/skills/skill-install-transaction.test.ts: installs, previews unchanged bytes, and updates a clean receipt-owned skill; preserves the old version after a %s filesystem failure. src/main/speech/model-manager-download-resume.test.ts: requests the remaining bytes when a clean range response ends before the file total; bounds a server that advances by pathologically tiny segments forever. src/main/ssh/config/ssh-config-parser.test.ts: parses a basic host block; parses multiple host blocks; skips wildcard-only Host entries; skips Host entries with only pattern characters; parses IdentityFile with ~ expansion; parses Windows-style IdentityFile with ~ expansion; parses quoted scalar values like OpenSSH; parses equals-form scalar directives; parses IdentityAgent with ~ expansion; parses IdentitiesOnly; parses GSSAPIAuthentication; keeps the first GSSAPIAuthentication value like OpenSSH; parses ProxyCommand, ProxyUseFdpass, and ProxyJump; ignores comments and blank lines; handles case-insensitive keywords; stops current block on Match directive; returns empty array for empty input; creates one parsed host per concrete alias on a multi-pattern Host line; applies identity agent settings to every concrete alias on a multi-pattern Host line; defaults port to 22 for invalid port values; converts hosts to SshTarget objects; uses host alias as hostname when HostName is missing; skips hosts that are already imported; defaults username to empty string when not specified; carries through identityFile, identityAgent, identitiesOnly, proxyCommand, and jumpHost; carries through gssapiAuthentication; imports duplicate aliases only once and keeps the first concrete host; maps hosts for the picker and flags aliases already in Orca; de-dupes aliases and defaults missing fields; parses hostname, user, port from ssh -G output; collects multiple identity files; parses forwardagent yes; defaults port to 22 when missing; returns empty hostname when missing; returns undefined user when missing; handles empty output; skips lines without spaces; parses proxycommand and filters \"none\"; parses proxyusefdpass yes; parses proxyjump and filters \"none\"; handles ~ expansion in identity file paths; handles Windows-style ~ expansion in identity file paths; parses identityagent with ~ expansion; preserves identityagent none so auth can disable agent fallback; parses identitiesonly yes; parses gssapiauthentication; parses controlmaster options and filters controlpath none. src/main/ssh/config/ssh-known-hosts-source.test.ts: splits the space-separated list ssh -G prints on one line; expands ~ in a reported path; keeps a double-quoted path containing spaces whole; falls back to the default files when ssh -G reported nothing; drops an explicit none without falling back to the defaults; splits both known_hosts lists on one line each; reads StrictHostKeyChecking, including accept-new; reads HostKeyAlias and prefers it for the lookup; reads the non-default port; reports that the lookup came from an alias, so the port is not appended; reports no alias when the dialed host is used; unions the entries of every file; skips a path that is a directory; returns nothing when no file can be read; reports nothing unreadable when every file parses; does not count an empty file as unreadable; prefers HostKeyAlias over the resolved hostname; uses the dialed host, because a resolved hostname can just echo the label; still prefers an explicit HostKeyAlias over the dialed host; falls back to the dialed host when nothing was resolved; is rejoined when the split fragments resolve to nothing; leaves a single unresolvable path alone rather than inventing one; expands the __PROGRAMDATA__ token rather than passing it through as a path; does not rewrite a path that merely BEGINS with the token characters; keeps a drive-letter path with mixed separators intact; splits a spaced Windows home into the fragments the rejoin has to put back. src/main/ssh/config/ssh-known-hosts.test.ts: reads the algorithm from the blob rather than trusting the line; rejects %s; refuses a malformed blob: %s; uses the bare host on the default port; tries the bracketed form first and falls back to bare off-port; matches an exact line; matches case-insensitively; reports a changed key of the same type; reports unknown for %s; matches any host in a comma list; matches the bracketed form on its port; does not apply a bracketed entry to the default port; falls back to a bare line when off-port; never reports a change from the fallback pass; matches a bare literal on the default port; matches a bracketed literal on its port; reports a change for a literal; handles %s; still reports a change for a different key on a revoked host; skips a line carrying an unrecognised marker; does not report a change when only another key type is on file; matches within the same type; accepts CRLF terminators; skips blanks, comments and surrounding whitespace; accepts a hit regardless of which file holds it (%s); never brackets the alias, whatever the port; still brackets an ordinary host on a non-default port; matches a bare alias entry on a non-default port; still accepts the unmodified key field; accepts a hashed entry whose salt is a full SHA1 digest; still accepts the unmutated ssh-keygen -H vector; falls back to the bare host line on a non-default port, as ssh did; refuses a key type it has not seen for a host it knows, as ssh did. src/main/ssh/connection/ssh-connect-config.test.ts: returns undefined when no default keys exist; returns the first existing key file; probes regular and FIDO2 keys in stable default order; keeps a regular default ahead of a malformed FIDO2 default; uses target host/port/username; falls back to resolved config when target fields are empty; keeps imported endpoint fields as the fallback when ssh -G is unavailable; sets readyTimeout to CONNECT_TIMEOUT_MS; sets keepaliveInterval to 15s; enables keyboard-interactive auth so MFA challenges can be answered; uses agent auth when no explicit key and SSH_AUTH_SOCK is set; enables agent forwarding when OpenSSH config requests it and an agent is available; does not enable agent forwarding without a usable agent; resolves IdentityAgent SSH_AUTH_SOCK from the environment; uses the Windows OpenSSH agent pipe when no environment socket is available on Windows; wraps agent auth with IdentityFile filtering when IdentitiesOnly is enabled; applies a manual target IdentitiesOnly setting when ssh -G reports it off; includes unencrypted target.identityFile auth when an agent is available; defers encrypted target.identityFile auth when an agent is available; uses keyFile auth when target.identityFile is set and no agent is available; includes unencrypted resolved identityFile auth when an agent is available; defers a default-path resolved identityFile to the no-agent retry; offers the keyboard-interactive challenge when the agent defers no key; provides fallback key when no agent is available; can force private key inclusion for the post-agent fallback path; falls back to resolved proxyCommand; returns structured jump-host config for target.jumpHost; returns structured jump-host config for resolved proxyJump; returns undefined when no proxy is configured. src/main/ssh/connection/ssh-connection-auth-fallback.test.ts: resolves OpenSSH config using configHost when present; tries ssh-agent before reading an explicit private key; falls back to direct private key auth when agent auth fails; falls back to direct private key auth after too many agent authentication failures; answers bounded keyboard-interactive challenges such as Duo 2FA. src/main/ssh/connection/ssh-connection-store.test.ts: lists picker suppression aliases without consulting re-adoption tombstones; getTarget delegates to store; addTarget generates an id and persists; saves the CLI control switch as given, with no consent stamp; addTarget stamps source as manual by default; addTarget preserves an explicitly provided source; inserts a new config host stamped as ssh-config; asks the parser for all hosts — reconciliation happens in the store; updates an existing config-sourced target when the port changed; refreshes gssapiAuthentication on sync; never overwrites a manual target that owns the alias; does not rewrite an unchanged config-sourced target; returns empty array when nothing changed; tombstones a deleted config-sourced host so re-import skips it; suppresses a deleted manual target from config discovery; re-adding a deleted host reclaims its alias so sync stops suppressing it; editing a target reclaims its alias from the deleted set; records a tombstone when removing any user-facing target; re-adopts orphaned repos when the same host is re-added. src/main/ssh/connection/ssh-connection.test.ts: transitions to connected on successful connect; captures the negotiated SSH server key fingerprint; allows concurrent exec commands for ssh2 transport; removes startup listeners after ssh2 connect succeeds; scopes lifecycle events and pending handshake timers to one mock client; enables TCP_NODELAY on the new ssh2 client after a reconnect cycle; transitions through connecting → connected states; reports error state on connection failure; disconnect cleans up and sets state to disconnected; getTarget returns a copy of the target; getState returns a copy of the state; throws when connecting a disposed connection. src/main/ssh/orcad-migration-stuck-host-profiles.test.ts: never leaves a running terminal behind; is offered only when leaving saved state behind would clear the refusal. src/main/ssh/orcad-runtime-maintenance.test.ts: keeps a connect from retrying a build an explicit Update also failed; offers Force when Update was refused by policy; forwards anew after an activation, since the new slot may bind another port; passes a refusal through and leaves the server linked; forwards anew once a serving slot is restored, which may have bound another port. src/main/ssh/port-forwarding/ssh-port-forward.test.ts: reports the port the OS bound when an ssh2 forward asks for port 0; adds an ssh2 port forward and returns entry; adds a system SSH port forward when no ssh2 client is available; reports the port a system SSH forward resolved when asked for port 0; passes resolved OpenSSH config to system SSH port forwards; disposes a system SSH tunnel when removing the forward; lists forwards filtered by connectionId; removes a forward by id; awaits forward close when removing a forward for user actions; returns null when removing nonexistent forward; removes all forwards for a connection; dispose removes all forwards; stores label in the entry; preserves the existing id when updating a forward succeeds. src/main/ssh/ssh-channel-multiplexer.test.ts: sends a JSON-RPC request and resolves on response; assigns unique request IDs; sends notifications without expecting a response; dispatches incoming notifications to handler; sends one keepalive frame per cadence tick; drives keepalive sends AND dead-link detection from a single interval; resolves false immediately on a disposed mux; throws on request after dispose; reports isDisposed correctly; clears registered handlers on dispose; does not retain handlers registered after dispose. src/main/ssh/system-transport/system-ssh-forward-process.test.ts: spawns exactly the argv the orphan matcher is given, LogLevel=INFO included; runs port forwards unmultiplexed even when ssh_config disables muxing; preserves manual target port and identity options in the forwarded ssh command; does not spawn ssh when the requested local forward port is already in use; spawns ssh after the requested local forward port preflight succeeds; resolves startup after the grace period when ssh keeps running; sends SIGTERM then SIGKILL when the process does not exit; does not resolve stop until the process exits. src/main/ssh/system-transport/system-ssh-loopback-route.test.ts: dials the held forward ahead of the alias so OpenSSH keeps these over ssh_config; drops a manual target's own port and proxies, which would fight the route; never routes a forward, whose argv must stay stable for orphan recovery; %s → %s; opens the forward over the real proxy, then proves a login over the route; lands the forward on the sshd and known_hosts name the direct route resolved; holds the forward with server-alive probes; never lets the background forward or probe ask for credentials; starts no forward when admission refuses the route; stops the forward when sshd refuses logins over it; stops the whole forward process tree on Windows, once; reports a forward that already exited before anyone listened for it. src/main/startup/browser-process-user-agent-ordering.test.ts: explains admission refusal before a desktop launch exits; explains an orcad holding the profile instead of exiting silently; does not acquire profile admission for a duplicate launch; executes after the dev app name and before later preflight work; exits and releases admission after pending profile move recovery fails; keeps Linux background startup failures console-only and immediate; keeps Linux serve startup failures console-only and immediate. src/main/startup/quit/quit-path-durable-write-blocking.test.ts: store.flushAsync() issues no synchronous fs syscalls; store.flushAsync() also moves the active-view and github-cache sidecars off the thread; stats.flushAsync() issues no synchronous fs syscalls; Grok session cleanup issues no synchronous fs syscalls; stats.flushAsync() closes out live agents before it returns, not when its write lands; a stalled mount leaves the event loop live instead of parking it; store.flushAsync() is one idempotent final barrier; store.flushAsync() resolves instead of throwing when the profile mount rejects writes. src/main/stats/agent-session-transition-recorder.test.ts: counts a hook-only agent that never writes an agent-shaped terminal title; never opens a session from a replayed or disk-restored working status; still closes a live session when the terminating status arrives as a replay; counts one session per turn across repeated working/done cycles; closes an open session when its pane is torn down; bounds the mirror and closes what it evicts; stops the session when the main agent settles and only a watch loop holds the row working; keeps the session open while a subagent outlives the main agent, and closes it when the child settles; dates the stop by the evidence when a shell outlives the last subagent; opens a new session dated by the observation when the main agent resumes after monitoring; never dates a reopen before the close a fact-less OSC repaint caused; dates the first live row after a restored one by the evidence, not the restored clock; treats an unchanged answer as a snapshot, never a transition; opens only on a live working edge; closes only a session it opened. src/main/text-generation/commit-message-text-generation-model-discovery.test.ts: returns static catalog models without spawning for static agents; discovers dynamic models through the agent CLI; discovers dynamic models through the configured agent command override; falls back to static models when dynamic discovery returns no parseable models; falls back to the first discovered non-Pi model when its static default is unavailable. src/main/text-generation/pull-request-context.test.ts: does not fetch unrelated fork remotes before generating PR context; bounds exact remote-ref probe concurrency for a large remote list; reports no branch change because PR context preparation is read-only; keeps a remote-qualified base when the selected base includes the remote. src/main/text-generation/source-control-local-process.test.ts: asks the supervisor to stop and leaves its group to it; signals nothing once the supervisor has exited; stops an agent that ignores its stdin end through its supervisor. src/main/updater/linux-package/linux-package-update-recovery.test.ts: retains the downloaded package with the digest from the event metadata; requires a string downloaded path; accepts an uppercase extension; requires a non-empty string version; matches an absolute file URL by basename; accepts duplicate entries that agree on the digest; refuses an entry with a missing or non-string digest; refuses a digest that is not base64 at all; refuses a malformed percent-encoded URL; refuses a missing or non-array file list; tolerates a non-object event; replaces the retained artifact with a newer download; keeps the retained artifact when a new download has no usable digest; keeps the retained artifact when an unrelated download event arrives; clears the retained artifact; preserves the artifact across a recheck of the same version; clears the artifact when another version takes over; preserves the artifact for an unknown or empty version; is a no-op without a retained artifact; returns the built command and package file name for a verified package; reports missing without a retained artifact; propagates a package-manager discovery failure; rejects a well-formed digest that belongs to another file; never hashes a package that a malformed digest failed to arm; rejects a path outside the cache root; rejects the cache root itself; rejects a package in the updater directory itself; rejects a pending directory that is a symlink out of the cache; performs one hash pass for concurrent requests; revalidates once the in-flight pass settles; does not reuse an in-flight pass for a different artifact; starts a fresh proof when the same package is captured again; returns the verified package path; returns missing without a retained artifact. src/main/updater/releases/updater-prerelease-feed-readiness.test.ts: accepts a GitHub %s asset redirect as ready; reports a GitHub asset request error as unavailable; reports not-ready with a verified last-good tag when the newest assets are unavailable; does not classify a non-positive tag limit as feed unavailability; reproduces the v1.4.142 publishing incident as not-ready; reports transport failures as unavailable instead of not-ready; treats an explicit asset 404 as not-ready when another asset is unavailable; accepts absolute manifest asset URLs without rewriting them to release asset paths; returns not-ready with an older ready update as last-good while newest is publishing. src/main/updater/releases/updater-release-builds.test.ts: lists hourly builds from the dedicated repo, newest first; lists daily builds from the dedicated repo, newest first; separates stable from rc in the shared main repo; skips drafts and unparseable tags; keeps a composed release title and drops one that repeats the tag; hides builds that published no artifact for this platform; returns an empty list when no build has this platform artifact; keeps mac builds visible on macOS regardless of the Windows leg; resolves the platform installer download url; leaves the installer url null when the release published no installer; tolerates a release whose assets are missing or malformed; sends the local gh token as a bearer header when one is available; sends no authorization header when gh has no token; retries unauthenticated once when GitHub rejects the token; does not retry a 401 that was already unauthenticated; falls back to the per-IP bucket when the token is rate limited and tells the gh breaker; reports a secondary limit carrying retry-after without retrying or tripping the breaker; records a primary limit in the token scope %s; keeps %s authenticated when only %s is blocked; skips the token while its scope %s is blocked; surfaces a rate limit with its reset time and a sign-in hint when unauthenticated; omits the sign-in hint when the rate-limited request was authenticated; treats 429 as a rate limit; reports a 403 without rate-limit headers as a plain HTTP error; reports a missing hourly repo distinctly; is null when GitHub sent no reset; prefers retry-after over the primary reset epoch; falls back to the primary reset epoch when there is no retry-after; reads retry-after as seconds; reads retry-after as an HTTP date; pins an hourly tag at the hourly repo download path; pins a daily tag at the daily repo download path; pins a stable tag at the main repo download path; rejects a tag that is not a version. src/main/updater/updater.build-channel-selection.test.ts: uses the display label in the unsupported-platform %s pinned-build error; still pins %s from an unsigned Windows dev build; restores ordinary release checks after local build selection fails; opts into the RC channel when checkForUpdatesFromMenu is called with includePrerelease; pins the generic feed to a perf-tagged prerelease when requested; surfaces no-update feedback when no newer perf-tagged prerelease exists; keeps prerelease publishing-window misses on the generic retry path; leaves the feed URL alone for a normal user-initiated check. src/main/updater/updater.check-preflight.test.ts: ignores a stale update-not-available event after a new check starts preflight; deduplicates repeated manual checks while the immediate checking status is active. src/main/updater/updater.prerelease-fallback.test.ts: retries a prerelease check once against the previous feed tag when the manifest is missing; surfaces a promise-only prerelease fallback failure after the primary error event. src/main/updater/updater.publishing-window-feed.test.ts: repins the generic feed to the newest RC tag for a prerelease user; repins the generic feed to a newer stable tag for a prerelease user; pins a stable fallback when Atom has no newer prerelease; uses last-good concrete feed when a user-initiated check lands during publishing; can offer an older fully published update while the newest release is publishing; keeps background publishing-window fallback on the short retry cadence. src/main/updater/updater.quit-and-install.test.ts: surfaces an accepted retry before electron-updater emits download progress; defers quitAndInstall and runs its launcher within the launch PATH scope; runs pre-quit cleanup before local PTY cleanup during update install; keeps optional pre-quit cleanup fail-and-continue behavior by default; ignores duplicate quitAndInstall requests while the shared delay is pending; arms the forced-exit watchdog once the install commits; disarms the forced-exit watchdog when sync install error recovery keeps the app open. src/main/window/createMainWindow-markdown-editor-focus.test.ts: ignores traffic light sync IPC on non-macOS; still intercepts Cmd+Shift+B and Cmd+Alt+B when the markdown editor is focused; coerces non-boolean setMarkdownEditorFocused payloads to false; opens a table-aware context menu synchronously without a renderer query; builds the plain edit menu for the one right-click the renderer marks as plain text. src/main/window/createMainWindow-renderer-crash-recovery.test.ts: notifies the caller when the renderer process is gone; passes the renderer webContents id through crash recording and recovery callbacks; forwards expected renderer teardowns so the recorder can diagnose suppression; reloads the app shell after an unexpected renderer process loss; does not reload after renderer loss when recovery is disabled; stops auto-reloading after a rapid renderer crash loop trips the breaker. src/main/window/dashboard-popout-window.test.ts: updates the title after language initialization and unsubscribes when closed; lets its preload expose saved WSL hosts in serve-only, and passes nothing otherwise; creates a native-frame window with the shared preload and no webview surface; shows the window on ready-to-show; loads the prod file entry; loads the dev server URL when in dev; focuses the existing window instead of creating a second one; trusts only the live popout webContents; restores a minimized window when re-requested; restores valid persisted bounds; discards off-screen persisted bounds and falls back to defaults; persists bounds on resize after the debounce, guarding near-minimum sizes; keeps saving bounds after a vetoed quit and freezes them only on an allowed quit; closeDashboardPopout closes an open window; applies the persisted app zoom level on dom-ready; follows app zoom changes while open and unsubscribes on close; zoomDashboardPopoutIfFocused steps the popout zoom only while focused; handles the zoom-in chord via before-input-event and ignores other keys; handles mouse-wheel zoom requests outside before-input-event; respects zoom keybinding overrides for keyboard and mouse-wheel paths. src/main/window/dropped-path-preparation.test.ts: opens and stores each unique source once as drop size grows tenfold; passes ordinary paths without copying for %s; preserves ordinary successes and copied successes when another file fails; serializes copy requests from different owners; reports all lost paths without a shared reason when their failures differ. src/main/window/main-window-updater-deferred-setup.test.ts: releases all 64 completed fallback handles without repeating initialization; keeps setup after ready-to-show and releases its fallback afterward; preserves the 15-second fallback when first paint never arrives; preserves destroyed-window guard reads before and at the fallback deadline; keeps the exact %s error and releases its completed fallback; keeps the exact fallback %s error without retrying initialization; preserves pending manual setup after failed ready registration; keeps setup options, live store reads, nudge writes and required quit flush. src/main/worker-threads/worker-thread-request-queue.test.ts: posts one request at a time and in the order it was dispatched; fires a posted call at its own deadline; stops respawning and fails the rest of the queue once deaths hit the cap; clears the death count when a fresh burst starts from full idle. src/main/workspace-hooks/hooks-setup-runner-script.test.ts: builds Windows runners for newline-heavy scripts without line-array splitting; builds POSIX runners without regex-wide CRLF normalization; writes POSIX setup runners for shebang-declared scripts on native Windows paths; keeps batch setup scripts on cmd.exe when the terminal is Git Bash; hands a Git Bash pane a launch command MSYS cannot rewrite for a cmd runner; replays interpreter flags declared on the shebang line; preserves cmd.exe setup runner semantics for configured cmd users; keeps POSIX runner behavior on POSIX platforms; waits when either local or project policy requires completed setup; marks setup-runner terminals for the always-on credential guard; maps git-bash to POSIX setup launch metadata on Windows; falls back to the cmd runner when the git-bash setting has no installed Git Bash; keeps the cmd runner for a non-Git bash such as Cygwin; keeps the cmd runner for a bare bash whose flavor cannot be resolved; uses the POSIX runner for a bare bash that resolves to Git Bash; uses the POSIX runner for an extension-less Git Bash path; preserves the existing cmd runner for PowerShell terminals; preserves cmd setup compatibility when a Windows-host project has a WSL shell setting; classifies an installed explicit Git Bash executable as a POSIX runner; falls back to the cmd runner when the explicit Git Bash path no longer exists. src/main/workspaces/catalog/repo-git-remote-identity-enrichment.test.ts: schedules remote identity enrichment without blocking the caller; probes a drive project whose runtime is WSL %s; sweeps a local repo: %s; probes an SSH row that carries only executionHostId on its own host; keeps same-path rows on two different SSH hosts from sharing one backoff; coalesces concurrent probes for the same repo location; caches no-identity probes briefly so list calls do not retry every time; does not rewrite the no-remote marker on a later retry; resolves a settled no-remote repo once it gains a remote; does not re-probe a resolved identity before the refresh window elapses; re-probes a resolved identity after the refresh window elapses; bounds re-probes per sweep so a large repo list drains gradually; forgets a removed repo location so its deadline cannot outlive the repo; keeps a surviving repo backoff when a sibling repo is removed; probes again when the same location is re-added after its probe was retired; coalesces concurrent sweeps instead of running one pass per list call. src/main/workspaces/creation/preparation/worktree-create-preparation.test.ts: preserves create priority through claim probes on %s; starts the checkout only once the async workspace root resolves; still deduplicates when both callers await the same pending root lookup; namespaces native Windows preparation directories for long paths; deduplicates preparation for the same repo, base, runtime, and workspace root; does not spend a divergence walk when the base matches exactly; claims when the two sides spell the same ref differently; evicts a repo's own stale preparation before another repo's; still discards armed checkouts when the project runtime needs repair; routes preparation and finalization through the selected WSL runtime; retries stale cleanup after a transient listing failure; does not discard a detached worktree with caller-controlled preparation metadata; does not re-arm after an isolated create; re-arms a preparation once creates arrive in a burst; holds the re-arm checkout until the create runs the deferred thunk; starts one explicit prefetch after the create completes; does not re-arm when finalization failed; reports a pending create while a stale-cleanup scan is running. src/main/workspaces/removal/local-worktree-removal-recovery.test.ts: recovers Git for Windows partial filesystem deletion failures; finishes a clean non-force deletion after Git leaves a Windows directory behind; recovers localized Windows failures after Git already removed the registration; does not recover unrelated localized failures while Git still owns the row; finishes a macOS deletion Git deregistered before failing with Directory not empty; leaves an orphaned checkout to the orphan cleanup instead of deleting it; prunes and strictly verifies on the selected WSL host without deleting files or branches. src/main/workspaces/removal/worktree-removal-safety.test.ts: rejects deleting a worktree that contains another registered worktree; rejects POSIX home directories even when host homedir has a different path shape; refuses a home the host reported even when no path rule recognises it; refuses the Windows host's home directory from a POSIX client; refuses the Windows host's home directory even when the host reported nothing. src/main/wsl/discovery/distro-catalog.test.ts: retries asynchronous discovery after a transient wsl.exe failure; retries synchronous discovery after a transient wsl.exe failure; retries asynchronous discovery after wsl.exe reports no distros yet; retries synchronous discovery after wsl.exe reports no distros yet; bounds the wsl.exe spawn rate while no distros are installed; bounds the asynchronous spawn rate while no distros are installed; still caches a non-empty distro list for the process lifetime; re-probes a docker-desktop-only machine once a user distro can appear; backs off while the list keeps coming back empty; backs off after repeated distro-list failures; caps the empty-list backoff at five minutes; keeps reporting an empty result after it goes stale; keeps reporting a known distro list indefinitely; bounds cached homes while retaining the most recently used distros; still lists distros, and an excluded distro home probe fails like any other probe; gives the async home probe time for a cold boot instead of killing it at 5 s; retries availability after a probe timeout instead of latching false; bounds the probe rate while the failure window is open; caches a successful probe for the process lifetime; names an explicit spawn directory, so no deleted cwd can read as ENOENT; shares one wsl.exe spawn between concurrent probes; keeps reporting the last observed answer after it goes stale; holds a definitive failure far longer than a timeout when %s; backs off after repeated probe failures; re-probes a seeded retryable failure once its window lapses; re-probes availability once a distro list succeeds after %s; does not drop an availability failure for an empty distro list; keeps reporting unavailable off Windows without probing. src/main/wsl/discovery/wsl-distro-holder.test.ts: runs a busybox-safe loop through --exec and names the distro; spawns a detached-style holder, records it, and waits for the distro to run; a holder exiting with code 1 notes its distro may have stopped (E9); rejects a pid whose command line is not our holder for that distro; cleans up stale and corrupt records; keeps a holder and its record when release cannot verify it; reports unverifiable when the command line is unreadable even with a matching pid; release kills a verified holder and removes its record; release never kills a pid that is no longer our holder; reports a spawned holder exit so liveness can re-check. src/main/wsl/filesystem/wsl-guest-drive-mountinfo.test.ts: skips 9p shares that are not drvfs, and malformed lines. src/main/wsl/orcad/wsl-orcad-deploy-app-update.test.ts: plans the update on the census the running orcad reports, not an unknown one; forwards force to the update planner; does not serve the incumbent past a refusal that is not a wait; does not re-activate the build this host rolled back from; labels the launch with the build the adoption found serving, not the stale record; defers while another run holds the host fence; lets an explicit force deploy past both skips; names a newer host refusal apart from a rolled-back one, so only Force is offered; updates an older host as before, without asking to adopt it; never asks the incumbent when it already runs this build. src/renderer/src/components/agent/agent-launch-picker-acquisition.test.ts: keeps a completed listing usable when the host rejects the wait parameter. src/renderer/src/components/feature-tips/feature-tip-telemetry.test.ts: delivers to chat-default users in all targeting while holding out not-switched users. src/renderer/src/components/native-chat/composer/pickers/use-native-chat-side-chat-command.test.ts: completes the row on a pick, even when the command is already typed; ignores a pick of the row while side chats are unavailable. src/renderer/src/components/native-chat/media/images/native-chat-image-copy.test.ts: explains a too-large refusal that arrives through the IPC error prefix. src/renderer/src/components/native-chat/messages/actions/native-chat-fork-launch.test.ts: keeps the question with the caller and says once that it never went; tells the user to check the side chat when delivery is unconfirmed; reports a delivered question as sent, silently. src/renderer/src/components/native-chat/messages/actions/native-chat-side-chat-fork-point.test.ts: decides each of %i tool-only turns once per layout, and nothing without a listener. src/renderer/src/components/native-chat/messages/actions/use-native-chat-fork.test.tsx: places a side chat whose question was %s: %o. src/renderer/src/components/terminal-pane/provider-notices/agent-presence/agent-terminal-presence.test.ts: calls back once for %s; stops watching once unsubscribed. src/renderer/src/components/terminal-pane/pty-connection/emulation/pty-connection-mode2031-raw-subscription.spec.ts: records a mode 2031 subscribe from the raw chunk boundary without answering. src/renderer/src/components/terminal-pane/pty-connection/input/pty-connection-deferred-input-forwarding.spec.ts: routes terminal input through deferPtyInput so the host can withhold it; forwards terminal input directly when the host supplies no deferPtyInput. src/renderer/src/components/terminal-pane/pty-connection/notifications/pty-connection-terminal-bell-attention.spec.ts: forwards onBell to pane attention and notification policy; hands the bell to the attention module with its pane even when pane attention is off. src/renderer/src/components/terminal-pane/pty-connection/output-restore/pty-connection-dropped-output-snapshot-healing.spec.ts: repaints from the main-owned snapshot when main drops pending output at the cap. src/renderer/src/components/terminal-pane/pty-connection/output/pty-connection-fresh-native-follow-reset.spec.ts: retries a fresh-spawn native follow reset when renderer dimensions return. src/renderer/src/components/terminal-pane/pty-connection/spawn/unbound-pane-spawn-recovery.test.ts: requests a tab remount for an unbound spawn; leaves recovery to a remote transport that armed its own retry; requests a remount when the transport reports no recovery of its own; settles the spawn as failed before requesting a remount; prefers the restored leaf id when reporting the anomaly; falls back to the pane leaf id when no restored leaf exists. src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-foreground-agent.test.ts: remove only literal registration import/call pairs import { registerPtyConnectionDeferredInputForwardingTests } from '../input/pty-connection-deferred-input-forwarding.spec', registerPtyConnectionDeferredInputForwardingTests(); no additional authored credit. src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-hidden-replay.test.ts: structural support changed only; no new authored removal credit. src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-input-and-repaint.test.ts: remove only literal registration import/call pairs import { registerPtyConnectionFreshNativeFollowResetTests } from '../output/pty-connection-fresh-native-follow-reset.spec', registerPtyConnectionFreshNativeFollowResetTests(); no additional authored credit. src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-spawn-and-startup.test.ts: remove only literal registration import/call pairs import { registerPtyConnectionTerminalBellAttentionTests } from '../notifications/pty-connection-terminal-bell-attention.spec', registerPtyConnectionTerminalBellAttentionTests(); no additional authored credit. src/renderer/src/components/terminal-pane/pty-transport/delivery/pty-dispatcher-delivery-interest.test.ts: registers interest on the first sidecar and releases on the last unsubscribe; releases sidecar interest only once for repeated unsubscribes; does not let an eager pre-mount buffer defeat hidden delivery gating; keeps eager buffers outside delivery interest when the PTY exits before mount; lets a sidecar exclusively own interest while an eager buffer overlaps. src/renderer/src/components/terminal-pane/pty-transport/delivery/pty-dispatcher-push-reattach.test.ts: blackholed delivery reproduces the wedge: no handler dispatch, no ACK ever. src/renderer/src/components/terminal-pane/pty-transport/ipc/ipc-pty-opencode-model.test.ts: sends preferences to the execution host and attaches only to its returned PTY; preserves one operation ID across a lost reply and reconnect; preserves ordinary raw startup and captured-session attachment. src/renderer/src/components/terminal-pane/pty-transport/output/pty-side-effect-pending-census.test.ts: sums registered gauges and drops disposed ones; ignores a repeated dispose instead of dropping another gauge; stops tracking new gauges past the cap so a missed dispose cannot grow it; tracks a live output processor queue through enqueue, drain, and dispose; excludes cleared consumed entries from retention after a bounded drain. src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-prebind-discard-notice.test.ts: says nothing once the pane is closed. src/renderer/src/components/terminal-pane/render-health/terminal-webgl-atlas-recovery.test.ts: refreshes after each scheduled atlas reset; falls back to a timeout when animation frames are unavailable; continues recovery when a manager throws after scheduling. src/renderer/src/components/terminal-pane/snapshots/pty-buffer-serializer.test.ts: tracks a remounted quiet title source even when the stale owner unregisters later; preserves the renderer-parsed output sequence in the response; preserves a pending escape tail for snapshot replay; routes a host's Reset Terminal request to the pane that owns the PTY. src/renderer/src/components/terminal-pane/snapshots/saved-terminal-output-replay.test.ts: marks panes with restored scrollback for fresh-shell viewport blanking. src/renderer/src/components/terminal-pane/viewport/split-right-white-screen.test.ts: recovers a 0x0 spawn whose container is briefly unmeasurable then settles. src/renderer/src/components/terminal-pane/viewport/visibility/stale-document-visibility.test.ts: removes global listeners when the last pane unregisters. src/renderer/src/components/ui/security-surface-theme-contrast.test.tsx: keeps trust text readable on a host-owned dialog surface; applies the hostile surface to ordinary dialogs, so the case above is exercised. src/renderer/src/components/use-terminal-parking-pass.test.tsx: does not queue a render when a pass produces the same parking sets; keeps parking transitions working under StrictMode. src/renderer/src/hooks/client-workspace-view-report.test.ts: reports an away user’s screen with no presence, and renews as soon as input returns; retries a failed send, empty or not, no sooner than a refresh. src/renderer/src/hooks/ipc-events/terminal-request-ipc-bridge-placement.test.ts: is new work in a workspace with no live terminal: its first bind moves it up in Recent. src/renderer/src/hooks/use-bundled-skill-install.test.tsx: ignores a repeated click while the install runs; retries the last requested skills on this host. src/renderer/src/hooks/useGitHubSlugMetadata.assignee-reuse.test.tsx: adds seeds to a successful empty list but not to loading or failed results; retains seed projection after expiry and fetches fresh metadata for a new reader. src/renderer/src/hooks/useSettingsNavigationMetadata.language-switch.test.tsx: retranslates the settings nav labels when the UI language changes live. src/renderer/src/lib/codex-stale-pane-account-identity.test.ts: keeps the prompt when the roster read fails and every label collapses. src/renderer/src/lib/image-blob-png.test.ts: downscales a pasted image above the pixel cap; releases its temporary URL on a decode error; releases its temporary URL when PNG encoding fails or exceeds the byte cap. src/renderer/src/lib/react-commit-cascade-store-write-samples.test.ts: names callers above the frame the write came from; restores Error.stackTraceLimit when the capture throws; stops collecting after the sample cap when captureStackTrace is unavailable; caps reported changed keys from a single wide write. src/renderer/src/lib/work-items/linear/linear-issue-context-snapshot.test.ts: applies the rendered text hard cap last and reserves the final overflow marker. src/renderer/src/runtime/rpc/targets/use-worktree-runtime-target.test.ts: keeps the runtime target identity stable across unrelated store writes. src/renderer/src/runtime/source-control/git/runtime-git-branch-client.test.ts: keeps creation available on old hosts and explains the missing operation; preserves landed warnings and direct collision recovery data; propagates remote discovery failure instead of a false empty page. src/renderer/src/runtime/sync-runtime-graph/sync-runtime-graph-key-reuse.test.ts: reuses unchanged worktree tab projections when one worktree title changes; bumps only the changed worktree version when a sibling stays byte-identical. src/renderer/src/store/repos/repos-catalog-request-coordination.test.ts: shares safe-auto fork sync cooldown attempts across store instances. src/renderer/src/store/slices/tabs/hydrated-workspace-reconciliation-batch.test.ts: collapses a 193-workspace hydration to one store write; leaves the store byte-identical to the per-workspace path; re-reconciling the batched result is a no-op, as it is for the per-workspace path. src/renderer/src/store/slices/tabs/hydrated-workspace-reconciliation-fixture.ts: dedicated zero-case helper retirement; no authored test credit. src/renderer/src/store/slices/worktrees/metadata/update-worktrees-meta.test.ts: persists each batch metadata update to its workspace owner. src/renderer/src/store/slices/worktrees/metadata/worktree-meta-rename.test.ts: dismisses a rename error on its capable remote owner; clears pending first-agent rename when the title is updated. src/renderer/src/store/terminals/store-sleep-shutdown-order.test.ts: orders capture, guards, handlers, main's stop, commit, cleanup, and exit settlement. src/renderer/src/store/terminals/terminal-first-tab-publication.test.ts: publishes the first terminal and root tab group atomically; publishes the first terminal and root tab group atomically, then marks it starting in main. src/renderer/src/store/terminals/workspace-terminal-reconnect-hydration.test.ts: restores daemon terminal wake hints before declaring the session ready. src/renderer/src/web/web-runtime-connection-auth-framing.test.ts: keeps legacy framing for sockets that cannot absorb a refused attempt. src/shared/terminal/appearance/terminal-font-weight-windows-regular.test.ts: moves the old Windows default to Regular once and queues the notice; keeps custom weights and fresh profiles without a notice; leaves macOS and Linux profiles untouched. tests/e2e/claude-agent-row-view.unit.test.ts: updates a mounted compact workflow description through the subscribed production parent; bounds upstream active-SSH visits for $population records, fallback=$fallback; derives %s destinations without requiring type or a workflow child ID; terminates a $path parent $status with no third request or intermediate toast; does not recover a first child %s. tests/e2e/cross-version-wire/agent-permission-legacy-spawn.unit.test.ts: keeps exact legacy ordering when structured creation cannot express afterTabId; can create a terminal without selecting the target worktree; preserves the legacy fresh-agent path when host authority is unavailable. tests/e2e/paired-instant-browser-tab.spec.ts: shows a paired browser tab before its create RPC resolves, then keeps it as one tab; adopts a paired browser tab without rebuilding its chrome. tests/e2e/session-tabs-decorative-title-fanout.unit.test.ts: bounds host publication and renderer application across remote worktrees; globally caps decorative freshness while every working status stays fresh. tests/e2e/session-tabs/mobile-session-active-performance.unit.test.ts: bounds saved-order publication selection at N and 10N without rematerializing navigation; bounds admitted placement reconciliation during supported reads at N and 10N; bounds selection input visits at 50 and 500 rows, including unavailable-target projection; keeps publication scoped to a changed worktree and shared by listeners at N and 10N; bounds preserved-chat merge and runtime-owned hydration at N and 10N (legacy=%s); bounds actual native retirement and shared delivery work at N and 10N. tests/e2e/session-tabs/mobile-session-active-publication.unit.test.ts: does not substitute a preserved row for an unavailable explicit legacy selection; resolves main-layout group selection %s (omitted group=%s). tests/e2e/structured-agent-session-launch-admission.unit.test.ts: opens a direct caller's terminal with the caller's own CLI arguments on a decline; opens the chat on the server that admitted it; opens only the terminal when this machine declines, without a notice; opens a new agent tab's terminal with the typed prompt when the caller names none; opens the chat, not a terminal, when this machine cannot resolve a new workspace yet; keeps its terminal fallback, as before. tests/e2e/structured-attention-prompt-relay.unit.test.ts: settles once per relayed prompt, not on every later status update. tests/e2e/structured-attention-read-retirement.unit.test.ts: Mark read on a chat never opened withdraws the prompt alert it surfaced; Mark read on a chat hidden since an earlier view covers the newer prompt it surfaced; reads a newly accepted visible prompt without sending an RPC for text-only updates. tests/e2e/structured-chat-tab-lifetime-cost.unit.test.ts: counts real raw stores, unrelated window publications and captured cleanup at %i chats; counts complete supplied-member sweep and separate final retirement at S=%i N=%i. tests/e2e/structured-session-workspace-authority.unit.test.ts: bounds actual target reveal across %i repo roots and metadata aliases. tests/e2e/terminal-agent-pane-launch.unit.test.ts: starts the host's planned command once, at the pane's size, and says it started; plans a pane whose prompt the window submits without the startup update prompt; preserves caller shell-ready delivery only for a connection-scoped adapter ($connectionId); starts no shell on plan failure (original failure: $returnLaunchFailure); cannot confirm a start whose provider answer was lost. tests/e2e/terminal-parked-memory.spec.ts: releases renderer terminal memory when hidden tabs park. tests/e2e/terminal-shortcuts.spec.ts: @headful Codex-like background output stays visible without disabling WebGL in auto mode; Shift with Russian layout text reaches the PTY as Cyrillic under kitty keyboard reporting. tests/e2e/terminal-split-activation-latency.spec.ts: records attributed CWD, spawn, bind, fixture-ready, input, and echo phases. tests/e2e/terminal-split-activation-latency-phases.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-split-activation-latency-main-probe.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-split-activation-latency-artifact.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-split-activation-latency-report.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-tab-switch-visual-restore.spec.ts: restores skipped hidden Grok output on light tab resume. tests/e2e/web-runtime-session-terminal-host-authority.unit.test.ts: delivers independent %s effort through the real prepared web callback; keeps same-ID local and runtime worktrees on the selected runtime owner; creates paired web agents through host authority so activation is mirrored; delivers generated continuation context after host-authoritative creation; seeds the chat composer for a draft that rode in on the launch command. Current changed source bytes, wrapper registrations and reduced selectors have no fresh execution or red/green credit. Unrelated surviving receipt objects keep their original narrow scope; current literal graph reachability does not establish historical import closure or equivalent replacement coverage. Current conservative command/import graph reaches src/main/persistence/worktree-sort-order-write.test.ts, src/main/runtime/construction/orca-runtime.test.ts, src/main/runtime/orca-runtime-tests/targeting-and-resilience.spec.ts, src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-04.spec.ts, src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-03.spec.ts, src/main/runtime/orca-runtime-tests/worktree-removal-and-reconciliation-part-02.spec.ts, src/main/runtime/orca-runtime-tests/gitlab-and-pr-bases-part-02.spec.ts, src/main/runtime/orca-runtime-tests/gitlab-and-pr-bases.spec.ts, src/main/runtime/orca-runtime-tests/worktree-setup-and-startup-part-03.spec.ts, src/main/runtime/orca-runtime-tests/worktree-setup-and-startup.spec.ts, src/main/runtime/orca-runtime-tests/terminal-sleep-and-teardown-part-02.spec.ts, src/main/runtime/orca-runtime-tests/mobile-summaries-part-03.spec.ts, src/main/runtime/orca-runtime-tests/mobile-summaries-part-02.spec.ts, src/main/runtime/orca-runtime-tests/mobile-summaries.spec.ts, src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-03.spec.ts, src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-02.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-13.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-12.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-11.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-10.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-08.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-07.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-05.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-04.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs-part-02.spec.ts, src/main/runtime/orca-runtime-tests/mobile-session-tabs.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-05.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-04.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-03.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status-part-02.spec.ts, src/main/runtime/orca-runtime-tests/terminal-handles-and-agent-status.spec.ts, src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-07.spec.ts, src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-04.spec.ts, src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-03.spec.ts, src/main/runtime/orca-runtime-tests/terminal-output-and-worker-recovery-part-02.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-12.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-11.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-10.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-09.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-08.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-07.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-06.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness-part-03.spec.ts, src/main/runtime/orca-runtime-tests/terminal-creation-and-readiness.spec.ts, src/main/runtime/orca-runtime-tests/headless-snapshots.spec.ts, src/main/runtime/orca-runtime-tests/terminal-side-effect-facts-part-02.spec.ts, src/main/runtime/orca-runtime-tests/local-worktree-creation-part-02.spec.ts, src/main/runtime/orca-runtime-tests/local-worktree-creation.spec.ts, src/main/runtime/integrations/skill/orca-runtime-skill-recovery.test.ts, src/main/runtime/orchestration/mailbox/routing/orchestration-mailbox-detached-routing.spec.ts, src/main/runtime/orchestration/mailbox/terminal-pointer/orchestration-mailbox-delivery.test.ts, src/main/runtime/orchestration/mailbox/terminal-pointer/orchestration-mailbox-pointer-cli-command.spec.ts, src/main/runtime/rpc/methods/terminal/terminal-multiplex-link-estimate.test.ts, src/main/runtime/rpc/methods/terminal/terminal-snapshot-scrollback-rows.test.ts, src/main/runtime/rpc/server/pairing/network/windows-default-route-interfaces.test.ts, src/main/runtime/rpc/server/pairing/pairing-offer-owner-cli.test.ts, src/main/runtime/rpc/terminal-multiplex-end-verdict.test.ts, src/main/runtime/session-graph/headless-runtime-graph.spec.ts, src/main/runtime/session-graph/runtime-graph-standing-contracts.test.ts, src/main/runtime/session-tabs/structured-conversation-tab-title.test.ts, src/main/runtime/session-tabs/terminal-close-observed-exit.test.ts, src/main/runtime/terminal/identity/terminal-identity-probe.test.ts, src/main/runtime/terminal/layout/paired-desktop-server-writes-layout-owner.test.ts, src/main/runtime/terminal/lifecycle/pty-lifecycle-generation-retention.test.ts, src/main/runtime/terminal/lifecycle/run-terminal-user-open.test.ts, src/main/runtime/terminal/surface/terminal-surface-replies-window.test.ts, src/main/runtime/terminal/surface/workspace-layout-projection.test.ts, src/main/runtime/terminal/view/terminal-subscribe-exit-waiter-leak.test.ts, src/main/runtime/terminal/wait/orca-runtime-exact-worker-provider-session.test.ts, src/main/runtime/terminal/wait/prime-agent-screen-readiness-transcripts.test.ts, src/main/runtime/workspaces/startup/runtime-worktree-agent-startup.test.ts, src/main/runtime/workspaces/startup/runtime-worktree-startup-readiness.test.ts, src/main/skills/skill-install-transaction.test.ts, src/main/speech/model-manager-download-resume.test.ts, src/main/ssh/config/ssh-config-parser.test.ts, src/main/ssh/config/ssh-known-hosts-source.test.ts, src/main/ssh/config/ssh-known-hosts.test.ts, src/main/ssh/connection/ssh-connect-config.test.ts, src/main/ssh/connection/ssh-connection-auth-fallback.test.ts, src/main/ssh/connection/ssh-connection-store.test.ts, src/main/ssh/connection/ssh-connection.test.ts, src/main/ssh/orcad-migration-stuck-host-profiles.test.ts, src/main/ssh/orcad-runtime-maintenance.test.ts, src/main/ssh/port-forwarding/ssh-port-forward.test.ts, src/main/ssh/ssh-channel-multiplexer.test.ts, src/main/ssh/system-transport/system-ssh-forward-process.test.ts, src/main/ssh/system-transport/system-ssh-loopback-route.test.ts, src/main/startup/browser-process-user-agent-ordering.test.ts, src/main/startup/quit/quit-path-durable-write-blocking.test.ts, src/main/stats/agent-session-transition-recorder.test.ts, src/main/text-generation/commit-message-text-generation-model-discovery.test.ts, src/main/text-generation/pull-request-context.test.ts, src/main/text-generation/source-control-local-process.test.ts, src/main/updater/linux-package/linux-package-update-recovery.test.ts, src/main/updater/releases/updater-prerelease-feed-readiness.test.ts, src/main/updater/releases/updater-release-builds.test.ts, src/main/updater/updater.build-channel-selection.test.ts, src/main/updater/updater.check-preflight.test.ts, src/main/updater/updater.prerelease-fallback.test.ts, src/main/updater/updater.publishing-window-feed.test.ts, src/main/updater/updater.quit-and-install.test.ts, src/main/window/createMainWindow-markdown-editor-focus.test.ts, src/main/window/createMainWindow-renderer-crash-recovery.test.ts, src/main/window/dashboard-popout-window.test.ts, src/main/window/dropped-path-preparation.test.ts, src/main/window/main-window-updater-deferred-setup.test.ts, src/main/worker-threads/worker-thread-request-queue.test.ts, src/main/workspace-hooks/hooks-setup-runner-script.test.ts, src/main/workspaces/catalog/repo-git-remote-identity-enrichment.test.ts, src/main/workspaces/creation/preparation/worktree-create-preparation.test.ts, src/main/workspaces/removal/local-worktree-removal-recovery.test.ts, src/main/workspaces/removal/worktree-removal-safety.test.ts, src/main/wsl/discovery/distro-catalog.test.ts, src/main/wsl/discovery/wsl-distro-holder.test.ts, src/main/wsl/filesystem/wsl-guest-drive-mountinfo.test.ts, src/main/wsl/orcad/wsl-orcad-deploy-app-update.test.ts, src/renderer/src/components/agent/agent-launch-picker-acquisition.test.ts, src/renderer/src/components/feature-tips/feature-tip-telemetry.test.ts, src/renderer/src/components/native-chat/composer/pickers/use-native-chat-side-chat-command.test.ts, src/renderer/src/components/native-chat/media/images/native-chat-image-copy.test.ts, src/renderer/src/components/native-chat/messages/actions/native-chat-fork-launch.test.ts, src/renderer/src/components/native-chat/messages/actions/native-chat-side-chat-fork-point.test.ts, src/renderer/src/components/native-chat/messages/actions/use-native-chat-fork.test.tsx, src/renderer/src/components/terminal-pane/provider-notices/agent-presence/agent-terminal-presence.test.ts, src/renderer/src/components/terminal-pane/pty-connection/emulation/pty-connection-mode2031-raw-subscription.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/input/pty-connection-deferred-input-forwarding.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/notifications/pty-connection-terminal-bell-attention.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/output-restore/pty-connection-dropped-output-snapshot-healing.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/output/pty-connection-fresh-native-follow-reset.spec.ts, src/renderer/src/components/terminal-pane/pty-connection/spawn/unbound-pane-spawn-recovery.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-foreground-agent.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-hidden-replay.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-input-and-repaint.test.ts, src/renderer/src/components/terminal-pane/pty-connection/tests/pty-connection-spawn-and-startup.test.ts, src/renderer/src/components/terminal-pane/pty-transport/delivery/pty-dispatcher-delivery-interest.test.ts, src/renderer/src/components/terminal-pane/pty-transport/delivery/pty-dispatcher-push-reattach.test.ts, src/renderer/src/components/terminal-pane/pty-transport/ipc/ipc-pty-opencode-model.test.ts, src/renderer/src/components/terminal-pane/pty-transport/output/pty-side-effect-pending-census.test.ts, src/renderer/src/components/terminal-pane/pty-transport/remote-runtime/remote-runtime-pty-transport-prebind-discard-notice.test.ts, src/renderer/src/components/terminal-pane/render-health/terminal-webgl-atlas-recovery.test.ts, src/renderer/src/components/terminal-pane/snapshots/pty-buffer-serializer.test.ts, src/renderer/src/components/terminal-pane/snapshots/saved-terminal-output-replay.test.ts, src/renderer/src/components/terminal-pane/viewport/split-right-white-screen.test.ts, src/renderer/src/components/terminal-pane/viewport/visibility/stale-document-visibility.test.ts, src/renderer/src/components/ui/security-surface-theme-contrast.test.tsx, src/renderer/src/components/use-terminal-parking-pass.test.tsx, src/renderer/src/hooks/client-workspace-view-report.test.ts, src/renderer/src/hooks/ipc-events/terminal-request-ipc-bridge-placement.test.ts, src/renderer/src/hooks/use-bundled-skill-install.test.tsx, src/renderer/src/hooks/useGitHubSlugMetadata.assignee-reuse.test.tsx, src/renderer/src/hooks/useSettingsNavigationMetadata.language-switch.test.tsx, src/renderer/src/lib/codex-stale-pane-account-identity.test.ts, src/renderer/src/lib/image-blob-png.test.ts, src/renderer/src/lib/react-commit-cascade-store-write-samples.test.ts, src/renderer/src/lib/work-items/linear/linear-issue-context-snapshot.test.ts, src/renderer/src/runtime/rpc/targets/use-worktree-runtime-target.test.ts, src/renderer/src/runtime/source-control/git/runtime-git-branch-client.test.ts, src/renderer/src/runtime/sync-runtime-graph/sync-runtime-graph-key-reuse.test.ts, src/renderer/src/store/repos/repos-catalog-request-coordination.test.ts, src/renderer/src/store/slices/tabs/hydrated-workspace-reconciliation-batch.test.ts, src/renderer/src/store/slices/tabs/hydrated-workspace-reconciliation-fixture.ts, src/renderer/src/store/slices/worktrees/metadata/update-worktrees-meta.test.ts, src/renderer/src/store/slices/worktrees/metadata/worktree-meta-rename.test.ts, src/renderer/src/store/terminals/store-sleep-shutdown-order.test.ts, src/renderer/src/store/terminals/terminal-first-tab-publication.test.ts, src/renderer/src/store/terminals/workspace-terminal-reconnect-hydration.test.ts, src/renderer/src/web/web-runtime-connection-auth-framing.test.ts, src/shared/terminal/appearance/terminal-font-weight-windows-regular.test.ts, tests/e2e/claude-agent-row-view.unit.test.ts, tests/e2e/cross-version-wire/agent-permission-legacy-spawn.unit.test.ts, tests/e2e/paired-instant-browser-tab.spec.ts, tests/e2e/session-tabs-decorative-title-fanout.unit.test.ts, tests/e2e/session-tabs/mobile-session-active-performance.unit.test.ts, tests/e2e/session-tabs/mobile-session-active-publication.unit.test.ts, tests/e2e/structured-agent-session-launch-admission.unit.test.ts, tests/e2e/structured-attention-prompt-relay.unit.test.ts, tests/e2e/structured-attention-read-retirement.unit.test.ts, tests/e2e/structured-chat-tab-lifetime-cost.unit.test.ts, tests/e2e/structured-session-workspace-authority.unit.test.ts, tests/e2e/terminal-agent-pane-launch.unit.test.ts, tests/e2e/terminal-parked-memory.spec.ts, tests/e2e/terminal-shortcuts.spec.ts, tests/e2e/terminal-split-activation-latency.spec.ts, tests/e2e/terminal-split-activation-latency-phases.ts, tests/e2e/terminal-split-activation-latency-main-probe.ts, tests/e2e/terminal-split-activation-latency-artifact.ts, tests/e2e/terminal-split-activation-latency-report.ts, tests/e2e/terminal-tab-switch-visual-restore.spec.ts, tests/e2e/web-runtime-session-terminal-host-authority.unit.test.ts. Unresolved local/computed edges remain explicit debt.", "Audit2095 current-parent preparation2126 (approved2092/2093 removals; source refresh2109/independent2112): Intentionally retire these low-value alarms without replacement: src/main/runtime/orca-runtime-tests/terminal-sleep-and-teardown-part-03.spec.ts: rejects exact terminal stop for multiple expected PTYs before stopping anything; omits stale renderer leaves when fresh PTY liveness is required; omits unbound renderer placeholders when fresh PTY liveness is required; captures the event method before the nested sleep receipt changes it; returns the close receipt without requiring a PTY stop; uses the same stop result as an empty explicit stop. src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-04.spec.ts: builds a compact worktree summary from persisted and live runtime state; reads the linked-PR state from the renderer repoId-keyed GitHub cache; carries persisted worktree host ownership in mobile summaries; emits only instance- and boundary-validated lineage parents in mobile summaries; resolves WSL platforms only for repos represented in mobile summaries; omits worktrees hidden by the host visibility policy from mobile summaries; applies linked-checkout source visibility to mobile summaries. src/main/runtime/orca-runtime-tests/hooks-and-hosted-review-part-02.spec.ts: routes runtime GitHub PR details and actions through the selected WSL project runtime; passes SSH connection context through hosted review creation flows; routes local WSL project hosted review flows through runtime git options; treats SSH worktree drift as unknown without local git probes. src/main/runtime/orca-runtime-tests/terminal-listing.spec.ts: emits one mobile session terminal tab per live PTY even if two tabs resolve to it; resolves projected split authority from the parent layout PTY binding; keeps explicit-id terminal lists from resolving all worktrees; keeps explicit-id cold-cache terminal lists from classifying unrelated same-repo worktrees; keeps unknown bare terminal-list ids on the no-scan exact-id path; matches explicit-id cwd PTYs for folder workspace instance IDs; routes PTY output through the PTY leaf index in large terminal graphs. src/main/runtime/orca-runtime-tests/window-authority.spec.ts: claims the first window as authoritative and ignores later windows; transfers authority from the headless sentinel to the first real window; relays terminal browser launches only while headless owns the graph; marks live headless PTYs for renderer reattach before desktop promotion; reports the activation gate state while no desktop window is available; bumps the epoch and enters reloading when the authoritative window reloads; can mark the graph ready for the authoritative window; drops back to unavailable and clears authority when the window disappears; hands the graph back to headless when a promoted serve window closes (#27292); keeps a crashed promoted serve renderer on its own window while the window is open; stays unavailable during initial loads before a graph is published. src/main/runtime/orchestration/mailbox/routing/orchestration-mailbox-routing-races.spec.ts: stops paged routing when the check is cancelled; keeps an unfiltered redrive unrestricted when reservations merge; keeps an unfiltered parked delivery unrestricted when reservations merge; uses a bounded indexed pane lookup for reminted Dispatch identity. src/main/runtime/workspaces/removal/worktree-teardown-unstopped-pty.test.ts: removes the reported wedged automation workspace without --force; clears PTY state once a failed stop is proven to have exited; names only the PTYs that are actually live, not every failed stop; warns with the specific sweep failure, not the generic deadline; still fails closed on a sweep-level failure without force. src/main/skills/bundled-skill-install/layouts.test.ts: links into a project only where its agent folder already exists; reads a pending cloud transaction as busy and a cloud receipt as another skill. src/main/skills/skill-upload-session-service.test.ts: joins concurrent disposal callers through exact cleanup completion. src/main/source-control/hosted-review-creation.test.ts: creates the pull request after fresh main-process validation passes; routes local WSL git and GitHub review creation through the selected runtime; creates a pull request on a GitHub Enterprise Server remote (#8312); creates a GitLab merge request after fresh main-process validation passes; creates an Azure DevOps pull request after fresh main-process validation passes; creates a Gitea pull request after fresh main-process validation passes. src/main/source-control/hosted-review-execution-host-routing.test.ts: reads a GitHub review on the SSH host a row names only in executionHostId; reads a GitLab review on the executionHostId host when connectionId names a different one; does not dial this client for a runtime row whose nested target shares a local name; keeps a runtime row with no nested target on this process (self-addressed stamp); keeps a legacy connectionId-only SSH row on its target; does not share a cached answer between two rows at one path on different hosts; creates a Bitbucket review locally for a runtime row rather than dialing its nested target; reports creation eligibility from the executionHostId-only SSH host; resolves the GitHub slug for a repo listing on the row it is asked about; refuses to dial a runtime host from this process; answers with the SSH target for both spellings and local otherwise; reads a runtime stamp on a row in this store as self-addressing, not a second machine. src/main/source-control/hosted-review-inflight-invalidation.test.ts: keeps other hosts, paths and their pending promises isolated. src/main/speech/model-manager.test.ts: requires pinned, internally consistent metadata for every model file; verifies downloaded model file hashes before installation; rejects non-HTTPS model downloads; marks OpenAI transcription models ready only when an API key is configured. src/main/ssh/connection/ssh-connection-sftp-namespace.test.ts: writes to the SFTP namespace path discovered on the same session; uploads the bundle into the SFTP namespace directory; issues no discovery requests when the caller supplies no mapping; ignores a mapping on a Windows host; never resolves on the system-SSH transport; leaves writeBuffer and openFileUploadSession on the shell path; keeps the shell path when discovery is inconclusive; preserves a session-limit channel-open error unchanged. src/main/ssh/connection/ssh-connection-system-transport.test.ts: uses system SSH transport when ProxyUseFdpass is resolved by OpenSSH; allows concurrent exec commands for system SSH with an Orca ControlMaster socket; keeps concurrent exec commands disabled for system SSH without a reusable socket; uses system SSH transport for ProxyCommand targets before ssh2 auth; keeps the original ssh2 reachability error when the system SSH probe fails; reports the system SSH failure when ssh2 cannot negotiate with the host; uses system transport for target or resolved OpenSSH proxy directives; allows an environment override for e2e coverage; ignores stale imported proxy fields when fresh OpenSSH config has no proxy. src/main/ssh/orcad-migration-catalog-client.test.ts: keeps each server project’s execution host, so a nested SSH row is never read as local; reads the server’s folder workspaces and the groups that place them on a host; sends and validates the %s operation; validates an abort result without letting it erase committed state; validates a committed receipt against the requested manifest; rejects malformed stage timestamps and abort envelopes; fails closed for method absence, malformed identity, and invalid pairing; refuses a destination without the capability before sending any migration request; reads method-not-found as the same refusal, not as a lost answer. src/main/ssh/runtime-ssh-access.test.ts: persists intent then claim before connecting and publishes redacted authenticated access; names the saved dependency when dependent state blocks an SSH access link; handles lost link responses without connecting or verifying again; invalidates old transport only after authenticated access is published; rejects changed request tuple after a completed link; rejects provisioning and missing durable generations without connecting; treats a lost successful unlink response as an idempotent no-op; hashes connection identity, not display state or owner; strictly validates ports and request IDs. src/main/startup/cli-launch-redirect.test.ts: detects a launch that received the unpacked CLI entrypoint; ignores normal desktop launches; ignores the entrypoint when it is only the executable itself (argv[0]); applies on Linux too; strips injected Chromium switches before node-mode CLI arguments; keeps user flags after the command and malformed boolean assignments; does not treat a later positional entrypoint path as the launcher; redirects a direct binary launch with no AppImage env at all; strips Chromium switches node mode would reject; preserves desktop-shaped switches after the command; leaves direct serve in-process but redirects its help; treats help as a CLI launch even without a command; treats %s as a CLI launch even without a command; does not treat a value of %s as a CLI early-exit flag; does not treat a serve option value as a help request; does not reinterpret help after the argument terminator; leaves a plain desktop launch alone; skips flag values when looking for the command positional; preserves a project selector in %j; does not apply the command form on macOS or Windows; never redirects an unpackaged build; runs the in-package CLI in Electron node mode with sanitized env; refuses to redirect twice so a dropped ELECTRON_RUN_AS_NODE cannot loop; reports a missing CLI entrypoint instead of booting the desktop app; execs a direct serve into the CLI so it keeps its pid and honours profile and supervise; keeps ORCA_USER_DATA_PATH for a direct serve without --user-data-dir; serves in-process when this runtime cannot exec or the exec fails; never execs the --serve child the CLI itself spawns, or serve on other platforms; surfaces a spawn failure as a non-zero exit; re-execs a direct --serve onto X11 when Electron resolved Wayland; leaves $name alone. src/main/startup/main-process-runtime-launch-activation.test.ts: marks startup settled for the launch record only after the window has loaded and settled; loads SSH targets on a windowless serve before paired clients can connect (#25886); pins the serve listener to loopback only for orcad browser sidecars (sidecar=$sidecar, #31895); saves Codex auth before a headless serve update restart, like the desktop does. src/main/startup/run-electron-vite-dev.test.ts: forwards dev instance identity to electron-vite; prepares userData orca and orca-dev wrappers for dev terminals; consumes the stable-name flag before forwarding args to electron-vite. src/main/updater/releases/updater-prerelease-feed.test.ts: returns the newest stable tag when the user is on an RC and a newer stable exists; returns the newest RC tag when no stable is newer than the current RC; can exclude prerelease tags for stable-channel checks; probes the %s %s platform manifest; returns null for stable-channel checks when only prereleases are newer; returns null when nothing in the feed is newer than the current version; ignores entries with unparseable tags; returns null when the fetch is not ok; returns null when the fetch throws; picks semver-newest across a mixed-order feed; ignores perf-tagged prereleases for regular RC checks; reports no RC update when only perf-tagged prereleases are newer; picks the semver-newest perf-tagged prerelease; matches only literal rc.N.perf prerelease tags for perf checks; reports no perf update instead of falling back to stable or RC tags; returns a bounded fallback candidate after the newest newer tag; returns null when the only newer tag has a manifest but its asset still 404s; requires a known old timestamp for a stable-fallback tag. src/main/updater/updater-events.test.ts: hands the whole downloaded event to artifact capture; publishes manual-install recovery after a %s download; publishes the normal downloaded state for AppImage builds; blocks downloaded-state handling when the packaged marker is unusable; passes the actual updater error into the install-failure handler; keeps manual-install recovery when a later check finds only the installed release; drops the artifact when another version takes over the cycle; ignores a downloaded event for an older target; accepts a matching event when the pending cache target was cleared; drops the artifact when progress reports a different pending version. src/main/updater/updater-linux-package-recovery-actions.test.ts: rejects both actions while no package-install recovery is active; validates the retained package on every invocation; restores recovery after a recheck resolves without a terminal event; restores recovery after a recheck fails; restores recovery when a pinned check resolves to the current version; restores recovery when resolving a pinned check fails; clears recovery for both actions once the package is gone; resolves %s as a result and keeps the card usable instead of rejecting; keeps recovery available after a transient read failure. src/main/updater/updater.check-settlement.test.ts: leaves a dismissed release update on the release source; surfaces net::ERR_FAILED to user-initiated checks with a friendly message; shows checking immediately for a user-initiated check while feed pinning is pending; keeps background checks event-driven before checking-for-update fires. src/main/usage/usage-provider-store-lifecycle.test.ts: persists identities for a whole first scan with one write; skips disabled and fresh matching states; invalidates prior sources on fingerprint changes and reuses them for forced scans. src/main/window/createMainWindow-close-confirmation.test.ts: clears the quit latch when the renderer prevents unload; does not notify the crash recorder when renderer teardown follows a confirmed window close; allows close when Electron reports a crashed webContents; requests confirmation for a hung-but-alive renderer instead of bypassing. src/main/workspaces/catalog/worktree-identity-persistence.test.ts: stores independent metadata for colliding locators on two hosts; projects canonical-only metadata for exactly one host; keeps canonical metadata current for legacy writers on a known owner; removes only the selected legacy-owner metadata when locators collide; keeps the canonical identity stable when the locator changes; leaves other hosts at the old locator when one host renames its folder. src/main/wsl/filesystem/wsl-drive-unc-conversion-startup.test.ts: leaves a journaled conversion untouched with the gate closed, %s. src/main/wsl/orcad/wsl-orcad-deploy.test.ts: holds the distro before it deploys and launches, and reports what C5 dials; passes its connect-state read (%s) on as hints; applies the SSH census and downgrade admission to the first deploy; refuses a WSL 1 distro before holding it or touching its files; tells an unregistered distro how to find the installed ones, before holding anything; surfaces an unsupported libc as the SSH host refusal does; starts an already-active slot the distro stop ended, instead of reporting it ready; deploys over an adopted unrecorded orcad under the SSH terminal gate, never a fresh launch; reports the adopted instance once the same version is serving; returns a deferral rather than launching over live terminals; hands a failed first deploy to the fresh-install cleanup before reporting it. src/main/zcode/zcode-v2-plan-credentials.test.ts: prefers migrated credentials over a stale legacy account; keeps legacy support when no personal v2 file exists; honors the upstream personal config override independently of the credential directory; does not borrow credentials from another execution host; preserves the protected Orca key priority even if v2 is corrupt; rejects upstream-invalid selection %j without legacy fallback; accepts a valid optional reasoning level; rejects upstream-invalid rule containers %j without legacy fallback; accepts a valid selected-provider empty personal rule and smart model rule; cannot establish selected credentials from %s rules; leaves encrypted CLI files unchanged across repeated account reads. tests/e2e/automation-terminal-launch.unit.test.ts: launches explicit Chat for %s through the common entry and keeps its actual fallback; does not apply stored Terminal extras to Chat; forwards only complete Chat-consumed values, with no Terminal startup base; refuses an expired admission without minting another launch identity; records the actual chat identity and delegates new workspace setup and provenance. tests/e2e/github-project-provider-columns.unit.test.ts: renders %s through one bounded items request on a named cloud host; preserves zero and complete counters: %j; preserves populated and legacy rows through shared sorting, grouping and cells. tests/e2e/structured-agent-session-launch-continuation.unit.test.ts: carries renderer lifetime through host intent to the real %s provider builders. tests/e2e/terminal-account-restart.unit.test.ts: replaces a tombstoned owner while its persisted pane binding has not yet been cleared; paired restart retains the old owner (host authority: %s). tests/e2e/terminal-hidden-view-parking.spec.ts: does not park excluded tabs; reproduces a static frame byte-for-byte across 25 park/reveal cycles. tests/e2e/terminal-long-table-scroll-restore.spec.ts: reproduces long markdown table artifacts after workspace switch and scroll; keeps narrow wrapped signer markdown table coherent after restore and scroll; keeps real emoji markdown table right edge clean after restore and scroll. tests/e2e/terminal-long-table-fixtures.ts: dedicated zero-case helper retirement; no authored test credit. tests/e2e/terminal-paste-ownership.spec.ts: Windows multiline keyboard paste normalizes terminal newlines with one PTY owner; right-click paste sends clipboard text to the focused terminal exactly once. tests/e2e/terminal-raw-emoji-table-scroll-restore.spec.ts: uses WebGL by default for ordinary terminal output when available @terminal-rendering-golden. tests/e2e/terminal-windows-shell-paste-ownership.spec.ts: cmd.exe default terminal keyboard paste preserves exact content with one PTY owner; Git Bash default terminal keyboard paste preserves POSIX-shaped content with one PTY owner; WSL terminal keyboard paste preserves Linux shell content with one PTY owner. tests/e2e/workspace-board-lane-virtualization.spec.ts: mounts a window of cards for a large lane and keeps lane indexes; renders later lane indexes after the lane scrolls. Current changed source bytes, wrapper registrations and reduced selectors have no fresh execution or red/green credit. Unrelated surviving receipt objects keep their original narrow scope; current literal graph reachability does not establish historical import closure or equivalent replacement coverage. Current conservative command/import graph reaches src/main/runtime/orca-runtime-tests/terminal-sleep-and-teardown-part-03.spec.ts, src/main/runtime/orca-runtime-tests/mobile-creation-and-orchestration-part-04.spec.ts, src/main/runtime/orca-runtime-tests/hooks-and-hosted-review-part-02.spec.ts, src/main/runtime/orca-runtime-tests/terminal-listing.spec.ts, src/main/runtime/orca-runtime-tests/window-authority.spec.ts, src/main/runtime/orchestration/mailbox/routing/orchestration-mailbox-routing-races.spec.ts, src/main/runtime/workspaces/removal/worktree-teardown-unstopped-pty.test.ts, src/main/skills/bundled-skill-install/layouts.test.ts, src/main/skills/skill-upload-session-service.test.ts, src/main/source-control/hosted-review-creation.test.ts, src/main/source-control/hosted-review-execution-host-routing.test.ts, src/main/source-control/hosted-review-inflight-invalidation.test.ts, src/main/speech/model-manager.test.ts, src/main/ssh/connection/ssh-connection-sftp-namespace.test.ts, src/main/ssh/connection/ssh-connection-system-transport.test.ts, src/main/ssh/orcad-migration-catalog-client.test.ts, src/main/ssh/runtime-ssh-access.test.ts, src/main/startup/cli-launch-redirect.test.ts, src/main/startup/main-process-runtime-launch-activation.test.ts, src/main/startup/run-electron-vite-dev.test.ts, src/main/updater/releases/updater-prerelease-feed.test.ts, src/main/updater/updater-events.test.ts, src/main/updater/updater-linux-package-recovery-actions.test.ts, src/main/updater/updater.check-settlement.test.ts, src/main/usage/usage-provider-store-lifecycle.test.ts, src/main/window/createMainWindow-close-confirmation.test.ts, src/main/workspaces/catalog/worktree-identity-persistence.test.ts, src/main/wsl/filesystem/wsl-drive-unc-conversion-startup.test.ts, src/main/wsl/orcad/wsl-orcad-deploy.test.ts, src/main/zcode/zcode-v2-plan-credentials.test.ts, tests/e2e/automation-terminal-launch.unit.test.ts, tests/e2e/github-project-provider-columns.unit.test.ts, tests/e2e/structured-agent-session-launch-continuation.unit.test.ts, tests/e2e/terminal-account-restart.unit.test.ts, tests/e2e/terminal-hidden-view-parking.spec.ts, tests/e2e/terminal-long-table-scroll-restore.spec.ts, tests/e2e/terminal-long-table-fixtures.ts, tests/e2e/terminal-paste-ownership.spec.ts, tests/e2e/terminal-raw-emoji-table-scroll-restore.spec.ts, tests/e2e/terminal-windows-shell-paste-ownership.spec.ts, tests/e2e/workspace-board-lane-virtualization.spec.ts. Unresolved local/computed edges remain explicit debt.", diff --git a/mobile/src/platform/host-os.web.test.ts b/mobile/src/platform/host-os.web.test.ts deleted file mode 100644 index 7673afefbbb..00000000000 --- a/mobile/src/platform/host-os.web.test.ts +++ /dev/null @@ -1,38 +0,0 @@ -// @vitest-environment happy-dom -import { afterEach, describe, expect, it } from 'vitest' -import { hostOs } from './host-os.web' - -function agent(userAgent: string, platform = '', maxTouchPoints = 0) { - Object.defineProperty(navigator, 'userAgent', { value: userAgent, configurable: true }) - Object.defineProperty(navigator, 'platform', { value: platform, configurable: true }) - Object.defineProperty(navigator, 'maxTouchPoints', { value: maxTouchPoints, configurable: true }) -} - -afterEach(() => agent('')) - -describe("the page's host OS", () => { - it('reads the phone off the WebView user agent', () => { - agent( - 'Mozilla/5.0 (iPhone; CPU iPhone OS 26_5 like Mac OS X) AppleWebKit/605.1.15 (KHTML, like Gecko) Mobile/15E148' - ) - expect(hostOs()).toBe('ios') - agent( - 'Mozilla/5.0 (Linux; Android 17; sdk_gphone64_arm64 Build/CP1A; wv) AppleWebKit/537.36 (KHTML, like Gecko) Version/4.0 Chrome/144.0 Mobile Safari/537.36' - ) - expect(hostOs()).toBe('android') - }) - - it('reads an iPad that still names itself in the user agent', () => { - agent( - 'Mozilla/5.0 (iPad; CPU OS 18_5 like Mac OS X) AppleWebKit/605.1.15 (KHTML, like Gecko) Mobile/15E148' - ) - expect(hostOs()).toBe('ios') - }) - - it('takes an iPad, which reports a Mac, by its touch points', () => { - agent('Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/605.1.15', 'MacIntel', 5) - expect(hostOs()).toBe('ios') - agent('Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/605.1.15', 'MacIntel', 0) - expect(hostOs()).toBe('macos') - }) -}) diff --git a/mobile/src/session/native-chat/composer/MobileNativeChatComposer.keyboard.web.test.tsx b/mobile/src/session/native-chat/composer/MobileNativeChatComposer.keyboard.web.test.tsx index 8a1960ed425..43bb2bc76f7 100644 --- a/mobile/src/session/native-chat/composer/MobileNativeChatComposer.keyboard.web.test.tsx +++ b/mobile/src/session/native-chat/composer/MobileNativeChatComposer.keyboard.web.test.tsx @@ -80,48 +80,6 @@ describe('mobile composer modifier submit in the paired web page', () => { return input } - async function render(props: Partial = {}) { - function Draft() { - const [value, setValue] = useState(props.value ?? ' hello ') - const composer = ( - 0} - getComposerEditGeneration={() => 0} - {...props} - value={value} - onChangeText={setValue} - onSend={onSend} - /> - ) - return props.editSession ? ( -
- {composer} -
- ) : ( - composer - ) - } - await act(async () => root.render()) - await act(async () => field().focus()) - } - - async function key(flags: KeyboardEventInit = {}) { - // Enter's base shape was captured in the native iOS shell; modifier flags are supplied, not recorded chords. - const event = new KeyboardEvent('keydown', { - key: 'Enter', - code: 'Enter', - keyCode: 13, - bubbles: true, - cancelable: true, - ...flags - }) - await act(async () => { - field().dispatchEvent(event) - }) - return event - } - async function selectEnd() { await act(async () => { const input = field() @@ -172,62 +130,6 @@ describe('mobile composer modifier submit in the paired web page', () => { expect(input.value).toBe('retained draft') }) - it.each([ - {}, - { shiftKey: true }, - { ctrlKey: true }, - { metaKey: true, shiftKey: true }, - { metaKey: true, altKey: true }, - { metaKey: true, ctrlKey: true }, - { metaKey: true, isComposing: true }, - { metaKey: true, keyCode: 229 } - ])('leaves editing and composition keys untouched: %j', async (flags) => { - await render() - expect((await key(flags)).defaultPrevented).toBe(false) - expect(onSend).not.toHaveBeenCalled() - expect(field().value).toBe(' hello ') - }) - - it('sends with Cmd+Enter, not Ctrl+Enter, in a desktop Mac browser', async () => { - vi.spyOn(navigator, 'userAgent', 'get').mockReturnValue( - 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/142.0.0.0 Safari/537.36' - ) - vi.spyOn(navigator, 'platform', 'get').mockReturnValue('MacIntel') - await render() - expect((await key({ ctrlKey: true })).defaultPrevented).toBe(false) - expect(onSend).not.toHaveBeenCalled() - expect((await key({ metaKey: true })).defaultPrevented).toBe(true) - expect(onSend).toHaveBeenCalledTimes(1) - }) - - it.each([ - { value: '@rea', expected: '@README.md ', filePaths: ['README.md', 'src/read.ts'] }, - { value: '/comp', expected: '/compact ', filePaths: [] } - ])( - 'completes the first existing menu row for $value without submitting', - async ({ value, expected, filePaths }) => { - await render({ - value, - filePaths, - agent: 'codex', - structured: { - catalog: { reported: [{ name: 'compact', kind: 'command' }], conversationCommands: [] }, - files: [], - onRemoveFile: vi.fn(), - onSlashMenuOpen: vi.fn() - } - }) - await selectEnd() - expect((await key({ metaKey: true })).defaultPrevented).toBe(true) - expect(field().value).toBe(expected) - expect(onSend).not.toHaveBeenCalled() - expect(document.activeElement).toBe(field()) - expect((await key({ metaKey: true, repeat: true })).defaultPrevented).toBe(true) - expect(onSend).not.toHaveBeenCalled() - expect(field().value).toBe(expected) - } - ) - function queuedGeometry() { const anchor = { top: 430, width: 390, clipTop: 250, clipHeight: 300 } vi.spyOn(HTMLElement.prototype, 'getBoundingClientRect').mockImplementation(function ( @@ -244,16 +146,6 @@ describe('mobile composer modifier submit in the paired web page', () => { return anchor } - async function queuedMention() { - await render({ - value: '@rea', - filePaths: ['README.md', 'src/read.ts'], - agent: 'codex', - editSession: { cancel: vi.fn(), refuseCommand: () => false, locked: false } - }) - await selectEnd() - } - function suggestion() { return container.querySelector('[aria-label="README.md"]') } @@ -339,59 +231,4 @@ describe('mobile composer modifier submit in the paired web page', () => { expect(suggestion()).toBeNull() expect(container.querySelector('[aria-label="queue.ts"]')).toBeNull() }) - - it('picks a suggestion outside the clipped queue while keeping the edited draft focused', async () => { - queuedGeometry() - await queuedMention() - const input = field() - const row = suggestion() - expect(row).not.toBeNull() - const queue = container.querySelector('[data-testid="queue-viewport"]')! - expect(queue.contains(row)).toBe(false) - expect(container.contains(row)).toBe(true) - const down = new MouseEvent('mousedown', { bubbles: true, cancelable: true }) - await act(async () => { - row!.dispatchEvent(down) - row!.dispatchEvent(new MouseEvent('mouseup', { bubbles: true })) - row!.click() - }) - expect(down.defaultPrevented).toBe(true) - expect(field()).toBe(input) - expect(input.value).toBe('@README.md ') - expect(document.activeElement).toBe(input) - expect(suggestion()).toBeNull() - expect(onSend).not.toHaveBeenCalled() - }) - - it('retires scrolled-out or hidden editor suggestions and restores them from current geometry', async () => { - const anchor = queuedGeometry() - await queuedMention() - expect( - suggestion(), - JSON.stringify({ - value: field().value, - caret: field().selectionStart, - focused: document.activeElement === field(), - viewport: window.innerHeight, - text: container.textContent?.slice(0, 120) - }) - ).not.toBeNull() - anchor.top = 900 - await act(async () => window.dispatchEvent(new Event('scroll'))) - expect(suggestion()).toBeNull() - anchor.top = 430 - await act(async () => window.dispatchEvent(new Event('scroll'))) - expect(suggestion()).not.toBeNull() - anchor.clipTop = 450 - anchor.clipHeight = 0 - await act(async () => window.dispatchEvent(new Event('resize'))) - expect(suggestion()).toBeNull() - anchor.clipTop = 250 - anchor.clipHeight = 300 - anchor.width = 0 - await act(async () => window.dispatchEvent(new Event('resize'))) - expect(suggestion()).toBeNull() - expect(field().value).toBe('@rea') - expect(document.activeElement).toBe(field()) - }) }) diff --git a/src/main/agent-hooks/server/testing/identity/server-kiro-program-status-identity.test.ts b/src/main/agent-hooks/server/testing/identity/server-kiro-program-status-identity.test.ts deleted file mode 100644 index f58869bff09..00000000000 --- a/src/main/agent-hooks/server/testing/identity/server-kiro-program-status-identity.test.ts +++ /dev/null @@ -1,64 +0,0 @@ -import { mkdtempSync, rmSync } from 'node:fs' -import { tmpdir } from 'node:os' -import { join } from 'node:path' -import { afterEach, beforeEach, expect, it } from 'vitest' -import { createProgramStatusProjector } from '../../../../../shared/program-status/agent-status-projector' -import { AgentHookServer } from '../../agent-hook-server' -import { buildBody, PANE, postHookEvent } from '../server.test-fixtures' - -const SESSION = 'sess_11111111-2222-4333-8444-555555555555' - -// Shape written by kiro-cli 3.1.0's TUI, which reports under its binary name. -const kiroOsc = (state: string): string => `\x1b]7501;state=${state}:app=kiro-cli\x07` - -let dir: string -let server: AgentHookServer -let projector: ReturnType - -beforeEach(async () => { - dir = mkdtempSync(join(tmpdir(), 'orca-kiro-program-status-')) - server = new AgentHookServer() - await server.start({ env: 'production', userDataPath: dir }) - projector = createProgramStatusProjector() -}) - -afterEach(() => { - server.stop() - rmSync(dir, { recursive: true, force: true }) -}) - -async function kiroHook(hookEventName: string, payload: Record = {}) { - const response = await postHookEvent( - server, - buildBody({ session_id: SESSION, hook_event_name: hookEventName, cwd: '/repo', ...payload }), - '/hook/kiro' - ) - expect(response.status).toBe(204) -} - -function kiroReports(state: string): void { - for (const { payload } of projector.process(kiroOsc(state)).payloads) { - server.ingestTerminalStatus({ - paneKey: PANE, - tabId: 'tab-1', - worktreeId: 'wt-1', - connectionId: null, - fromProgramStatus: true, - payload - }) - } -} - -function recordedRow() { - const row = server.getStatusSnapshot()[0] - return { agentType: row?.agentType, sessionId: row?.providerSession?.id } -} - -it('keeps the Kiro session through a turn that Kiro also reports over OSC 7501', async () => { - kiroReports('working') - await kiroHook('UserPromptSubmit', { prompt: 'hi' }) - await kiroHook('Stop') - kiroReports('done') - - expect(recordedRow()).toEqual({ agentType: 'kiro', sessionId: SESSION }) -}) diff --git a/src/main/agent-hooks/server/testing/providers/codex/server-codex-session-start.test.ts b/src/main/agent-hooks/server/testing/providers/codex/server-codex-session-start.test.ts index 31dd5bd0635..953728ce3f0 100644 --- a/src/main/agent-hooks/server/testing/providers/codex/server-codex-session-start.test.ts +++ b/src/main/agent-hooks/server/testing/providers/codex/server-codex-session-start.test.ts @@ -34,95 +34,86 @@ function previousRoot(children: number): AgentHookEventPayload { } } -it.each([10, 100])( - 'retires %i foreign-session children at a cached startup boundary', - (children) => { - const desktop = new AgentHookServer() - desktop.beginRemoteRelaySession('wsl:Ubuntu') - desktop.ingestRemote(previousRoot(children), 'wsl:Ubuntu') - expect(desktop.getStatusSnapshot()[0].subagents).toHaveLength(children) - if (children === 100) { - desktop.beginRemoteRelaySession('wsl:Ubuntu') - } - const incoming: AgentHookEventPayload = { - ...previousRoot(0), - toolAgentId: undefined, - hookEventName: 'SessionStart', - providerSession: { key: 'session_id', id: 'new-root' }, - payload: { - agentType: 'codex', - state: 'done', - sessionBoundary: true, - prompt: 'new context', - model: 'new-model', - mainAgent: { state: 'done', outcome: 'cancellation', stateStartedAt: 2000 } - } - } - desktop.ingestRemote( - buildCachedPaneReplayEnvelope({ event: incoming, meta: { source: 'codex' } }), - 'wsl:Ubuntu' - ) - expect(desktop.getStatusSnapshot()[0]).toMatchObject({ +it('retires foreign-session children at a cached startup boundary', () => { + const children = 100 + const desktop = new AgentHookServer() + desktop.beginRemoteRelaySession('wsl:Ubuntu') + desktop.ingestRemote(previousRoot(children), 'wsl:Ubuntu') + expect(desktop.getStatusSnapshot()[0].subagents).toHaveLength(children) + desktop.beginRemoteRelaySession('wsl:Ubuntu') + const incoming: AgentHookEventPayload = { + ...previousRoot(0), + toolAgentId: undefined, + hookEventName: 'SessionStart', + providerSession: { key: 'session_id', id: 'new-root' }, + payload: { + agentType: 'codex', state: 'done', - restoredUnconfirmed: true, - providerSession: { id: 'new-root' }, + sessionBoundary: true, prompt: 'new context', model: 'new-model', mainAgent: { state: 'done', outcome: 'cancellation', stateStartedAt: 2000 } - }) - expect(desktop.getStatusSnapshot()[0].subagents).toBeUndefined() - expect(desktop._getStateForTests().codexSubagentRosterByPaneKey.get(PANE)?.size).toBe(0) - desktop.ingestRemote({ ...incoming, hookEventName: undefined }, 'wsl:Ubuntu') - expect(desktop.getStatusSnapshot()[0].subagents).toBeUndefined() - } -) - -it.each(['previous', 'incoming'] as const)( - 'preserves live context when the %s session ID is missing', - (missing) => { - const desktop = new AgentHookServer() - const previous = previousRoot(1) - vi.useFakeTimers({ toFake: ['Date'] }) - vi.setSystemTime(1_000) - try { - desktop.ingestRemote( - { - ...previous, - hookEventName: 'UserPromptSubmit', - toolAgentId: undefined, - hasExplicitPrompt: true, - providerSession: missing === 'previous' ? undefined : previous.providerSession, - payload: { ...previous.payload, mainAgent: { state: 'working', stateStartedAt: 1000 } } - }, - 'wsl:Ubuntu' - ) - expect(desktop.getStatusSnapshot()[0].turnStartedAt).toBe(1_000) - vi.setSystemTime(3_000) - desktop.ingestRemote( - { - ...previous, - toolAgentId: undefined, - hookEventName: 'SessionStart', - providerSession: - missing === 'incoming' ? undefined : { key: 'session_id', id: 'new-root' }, - payload: { agentType: 'codex', state: 'working', prompt: '' } - }, - 'wsl:Ubuntu' - ) - expect(desktop.getStatusSnapshot()[0]).toMatchObject({ - state: 'working', - prompt: 'old task', - model: 'old-model', - mainAgent: { state: 'working', stateStartedAt: 1000 }, - subagents: [{ id: 'old-0', state: 'working' }], - turnStartedAt: 1_000 - }) - } finally { - desktop.stop() - vi.useRealTimers() } } -) + desktop.ingestRemote( + buildCachedPaneReplayEnvelope({ event: incoming, meta: { source: 'codex' } }), + 'wsl:Ubuntu' + ) + expect(desktop.getStatusSnapshot()[0]).toMatchObject({ + state: 'done', + restoredUnconfirmed: true, + providerSession: { id: 'new-root' }, + prompt: 'new context', + model: 'new-model', + mainAgent: { state: 'done', outcome: 'cancellation', stateStartedAt: 2000 } + }) + expect(desktop.getStatusSnapshot()[0].subagents).toBeUndefined() + desktop.ingestRemote({ ...incoming, hookEventName: undefined }, 'wsl:Ubuntu') + expect(desktop.getStatusSnapshot()[0].subagents).toBeUndefined() +}) + +it('preserves live context when the incoming session ID is missing', () => { + const desktop = new AgentHookServer() + const previous = previousRoot(1) + vi.useFakeTimers({ toFake: ['Date'] }) + vi.setSystemTime(1_000) + try { + desktop.ingestRemote( + { + ...previous, + hookEventName: 'UserPromptSubmit', + toolAgentId: undefined, + hasExplicitPrompt: true, + providerSession: previous.providerSession, + payload: { ...previous.payload, mainAgent: { state: 'working', stateStartedAt: 1000 } } + }, + 'wsl:Ubuntu' + ) + expect(desktop.getStatusSnapshot()[0].turnStartedAt).toBe(1_000) + vi.setSystemTime(3_000) + desktop.ingestRemote( + { + ...previous, + toolAgentId: undefined, + hookEventName: 'SessionStart', + providerSession: undefined, + payload: { agentType: 'codex', state: 'working', prompt: '' } + }, + 'wsl:Ubuntu' + ) + expect(desktop.getStatusSnapshot()[0]).toMatchObject({ + state: 'working', + prompt: 'old task', + model: 'old-model', + mainAgent: { state: 'working', stateStartedAt: 1000 }, + subagents: [{ id: 'old-0', state: 'working' }], + turnStartedAt: 1_000 + }) + } finally { + desktop.stop() + vi.useRealTimers() + } +}) it.each(['local', 'relay', 'old-relay', 'old-relay-no-main'] as const)( 'separates Codex session identity from live turns on %s', diff --git a/src/main/agent-hooks/server/testing/providers/opencode/server-opencode2-program-status.test.ts b/src/main/agent-hooks/server/testing/providers/opencode/server-opencode2-program-status.test.ts index 14f73339076..30f70b2a62d 100644 --- a/src/main/agent-hooks/server/testing/providers/opencode/server-opencode2-program-status.test.ts +++ b/src/main/agent-hooks/server/testing/providers/opencode/server-opencode2-program-status.test.ts @@ -1,7 +1,7 @@ import { afterEach, expect, it, vi } from 'vitest' import { AgentHookServer } from '../../../agent-hook-server' import { HOOK_SILENCE_BEFORE_PROGRAM_STATUS_FALLBACK_MS } from '../../../program-status/hook-precedence' -import { buildBody, ingestProgramStatusOsc, PANE, postHookEvent } from '../../server.test-fixtures' +import { buildBody, ingestProgramStatusOsc, postHookEvent } from '../../server.test-fixtures' vi.mock('../../../../../telemetry/client', () => ({ track: vi.fn() })) vi.mock('../../../../../telemetry/cohort-classifier', () => ({ getCohortAtEmit: () => ({}) })) @@ -12,42 +12,11 @@ afterEach(() => { server?.stop() }) -function oscStatus(state: 'working' | 'done'): void { - // OpenCode 2 reports OSC 7501 as `app=opencode`. - server!.ingestTerminalStatus({ - paneKey: PANE, - tabId: 'tab-1', - worktreeId: 'wt-1', - connectionId: null, - fromProgramStatus: true, - payload: { state, prompt: '', agentType: 'opencode' } - }) -} - async function hook(hook_event_name: string, prompt = ''): Promise { const body = buildBody({ hook_event_name, sessionID: 'ses_opencode2', prompt }) expect((await postHookEvent(server!, body, '/hook/opencode2')).status).toBe(204) } -it('keeps OpenCode 2 hooks and their session on a pane its OSC 7501 reports as opencode', async () => { - server = new AgentHookServer() - await server.start({ env: 'production' }) - oscStatus('working') - await hook('SessionBusy', 'rename the branch') - expect(server.getStatusSnapshot()[0]).toMatchObject({ - agentType: 'opencode2', - prompt: 'rename the branch' - }) - - await hook('SessionIdle') - oscStatus('done') - expect(server.getStatusSnapshot()[0]).toMatchObject({ - agentType: 'opencode2', - state: 'done', - providerSession: { id: 'ses_opencode2' } - }) -}) - it('keeps an idle OpenCode 2 pane done while another session in its folder works', async () => { server = new AgentHookServer() await server.start({ env: 'production' }) diff --git a/src/main/artifacts/artifact-cloud-config.test.ts b/src/main/artifacts/artifact-cloud-config.test.ts index 52c6752fcbc..945ab0e0216 100644 --- a/src/main/artifacts/artifact-cloud-config.test.ts +++ b/src/main/artifacts/artifact-cloud-config.test.ts @@ -1,30 +1,7 @@ import { describe, expect, it } from 'vitest' -import { - allowsArtifactCloudAuthOverride, - resolveArtifactCloudApiUrl -} from './artifact-cloud-config' +import { resolveArtifactCloudApiUrl } from './artifact-cloud-config' describe('resolveArtifactCloudApiUrl', () => { - it('allows loopback HTTP only in development', () => { - expect( - resolveArtifactCloudApiUrl( - undefined, - { ORCA_ARTIFACTS_API_URL: 'http://127.0.0.1:45961' }, - false - ) - ).toBe('http://127.0.0.1:45961') - expect(() => resolveArtifactCloudApiUrl('http://127.0.0.1:45961', {}, true)).toThrow(/HTTPS/) - }) - - it('rejects origins that could receive an Orca access token', () => { - expect(() => resolveArtifactCloudApiUrl('https://example.com', {}, false)).toThrow( - /onorca\.dev/ - ) - expect(() => resolveArtifactCloudApiUrl('https://share.onorca.dev/path', {}, false)).toThrow( - /origin/ - ) - }) - it('allows both first-party domain families without trusting similar hostnames', () => { for (const origin of ['https://share.onorca.dev', 'https://share.orca.dev']) { expect(resolveArtifactCloudApiUrl(origin, {}, true)).toBe(origin) @@ -38,10 +15,4 @@ describe('resolveArtifactCloudApiUrl', () => { expect(() => resolveArtifactCloudApiUrl(origin, {}, true)).toThrow() } }) - - it('allows auth token overrides only in non-production development builds', () => { - expect(allowsArtifactCloudAuthOverride({}, false)).toBe(true) - expect(allowsArtifactCloudAuthOverride({ NODE_ENV: 'production' }, false)).toBe(false) - expect(allowsArtifactCloudAuthOverride({}, true)).toBe(false) - }) }) diff --git a/src/main/claude/structured/claude-conversation-title-owner.test.ts b/src/main/claude/structured/claude-conversation-title-owner.test.ts deleted file mode 100644 index b6e29f72b51..00000000000 --- a/src/main/claude/structured/claude-conversation-title-owner.test.ts +++ /dev/null @@ -1,23 +0,0 @@ -import { afterEach, expect, it, vi } from 'vitest' -import { acquired, fakeClaude, PROVIDER_SESSION_ID } from './claude-structured-session-test-support' -import { agentSessionRecordFixture } from '../../../shared/agent-session/lifecycle/agent-session-record.test-fixture' - -afterEach(() => vi.restoreAllMocks()) - -it('retains retirement knowledge when the optional title method is missing', async () => { - const cli = fakeClaude() - const adapter = await acquired(cli) - try { - const record = agentSessionRecordFixture() - record.sessionId = 'session-1' - record.providerHandleChain[0]!.handle.nativeId = PROVIDER_SESSION_ID - const owner = adapter.captureConversationTitleOwner(record)! - expect(owner.isCurrent(record)).toBe(true) - expect(await owner.generate('Repair login', () => owner.isCurrent(record))).toBeNull() - await adapter.closeAll() - expect(owner.isCurrent(record)).toBe(false) - } finally { - await adapter.closeAll() - cli.dispose() - } -}) diff --git a/src/shared/app-version.test.ts b/src/shared/app-version.test.ts deleted file mode 100644 index fb599d3c5e7..00000000000 --- a/src/shared/app-version.test.ts +++ /dev/null @@ -1,14 +0,0 @@ -import { describe, expect, it } from 'vitest' -import { parseCliVersion } from './app-version' - -describe('parseCliVersion', () => { - it('reads a version that ends a sentence', () => { - expect( - parseCliVersion("GitHub Copilot CLI 1.0.109.\nRun 'copilot update' to check for updates.") - ).toBe('1.0.109') - }) - - it('rejects a four-part version', () => { - expect(parseCliVersion('tool 1.2.3.4')).toBeNull() - }) -}) diff --git a/src/shared/speech-model-suggestion.test.ts b/src/shared/speech-model-suggestion.test.ts deleted file mode 100644 index abb0e6e5109..00000000000 --- a/src/shared/speech-model-suggestion.test.ts +++ /dev/null @@ -1,44 +0,0 @@ -import { describe, expect, it } from 'vitest' -import { suggestSpeechModel } from './speech-model-suggestion' -import type { SpeechModelErrorRate, SpeechModelManifest } from './speech-types' - -const rate = (language: string, percent: number): SpeechModelErrorRate => ({ - language, - percent, - metric: 'wer', - test: 'fleurs', - clips: 5, - measuredOn: '2026-11-09' -}) - -const model = (id: string, extra: Partial): SpeechModelManifest => ({ - id, - label: id, - type: 'transducer', - provider: 'local', - language: 'multilingual', - languages: [], - sampleRate: 16000, - streaming: false, - ...extra -}) - -const catalog = [ - model('parakeet', { recommended: true, languages: ['en', 'fr'], errorRates: [rate('fr', 8)] }), - model('sense', { languages: ['en', 'ko'], suggestedFor: ['ko'], errorRates: [rate('ko', 10)] }), - model('nemotron', { languages: ['fr', 'ko', 'tr'], errorRates: [rate('tr', 14)] }), - model('turbo', { - languages: ['fr', 'ko', 'tr', 'cy'], - errorRates: [rate('fr', 5), rate('ko', 3), rate('tr', 8)] - }), - model('retired', { languages: ['tr'], errorRates: [rate('tr', 1)], replacedBy: 'turbo' }) -] - -describe('suggestSpeechModel', () => { - it('keeps the user’s own pick while it is offered', () => { - const pick = (selectedModelId: string) => - suggestSpeechModel(catalog, 'tr', { selectedModelId, favorAccuracy: false })?.id - expect(pick('nemotron')).toBe('nemotron') - expect(pick('retired')).toBe('turbo') - }) -}) From a78aada6b35d76de6cca96b38a16af9e7a9b02af Mon Sep 17 00:00:00 2001 From: Luoqiu1 <49295370+Luoqiu1@users.noreply.github.com> Date: Thu, 12 Nov 2026 11:28:54 +0800 Subject: [PATCH 11/13] fix(native-chat): refresh skills when reopening the picker (#41023) * fix(native-chat): refresh skills when reopening the picker * fix(native-chat): show the last skill list while the reopened picker rechecks Clearing the renderer snapshot on close made every / or $ open start in a loading state. Keep the last answer per pane, show it at once on reopen, and ask the host once per open. * fix(native-chat): keep a late skill scan as the next open's fallback * test(native-chat): feed skill picker tests fixed host answers Importing the main-process scanner pulled src/main into the web typecheck project. --------- Co-authored-by: Kelvin Amoaba <97001695+AmoabaKelvin@users.noreply.github.com> --- .../pickers/use-native-chat-skills.test.tsx | 98 +++++++++++++++++++ .../pickers/use-native-chat-skills.ts | 21 +++- 2 files changed, 116 insertions(+), 3 deletions(-) create mode 100644 src/renderer/src/components/native-chat/composer/pickers/use-native-chat-skills.test.tsx diff --git a/src/renderer/src/components/native-chat/composer/pickers/use-native-chat-skills.test.tsx b/src/renderer/src/components/native-chat/composer/pickers/use-native-chat-skills.test.tsx new file mode 100644 index 00000000000..5717d1b94b7 --- /dev/null +++ b/src/renderer/src/components/native-chat/composer/pickers/use-native-chat-skills.test.tsx @@ -0,0 +1,98 @@ +// @vitest-environment happy-dom +import { act, cleanup, renderHook, waitFor } from '@testing-library/react' +import { afterEach, beforeEach, expect, it, vi } from 'vitest' +import type { DiscoveredSkill, SkillDiscoveryResult } from '../../../../../../shared/skills' +import { useAppStore } from '@/store' +import { makeFolderWorkspace } from '@/store/folder-workspaces/folder-workspace.test-fixture' +import { useNativeChatSkills } from './use-native-chat-skills' + +const rpc = vi.hoisted(() => ({ discover: vi.fn() })) +vi.mock('@/runtime/rpc/calls/runtime-rpc-call', () => ({ callRuntimeRpc: rpc.discover })) +vi.mock('@/store', async () => { + const { createTestStore } = await import('@/store/testing/app-store-test-fixture') + return { useAppStore: createTestStore() } +}) + +const SKILLS_ROOT = '/home/user/.agents/skills' +let installed: string[] +let tabId: string + +beforeEach(() => { + installed = [] + useAppStore.setState(useAppStore.getInitialState(), true) + useAppStore.setState({ + folderWorkspaces: [makeFolderWorkspace({ id: 'skills', folderPath: '/home/user/project' })] + }) + tabId = useAppStore.getState().createUnifiedTab('folder:skills', 'agent-session', { + agentSessionAgent: 'codex', + executionHostId: 'local' + }).id + rpc.discover.mockReset().mockImplementation(async () => inventory()) +}) + +afterEach(() => { + cleanup() +}) + +function inventory(): SkillDiscoveryResult { + const skills = installed.map((name): DiscoveredSkill => ({ + id: name, + name, + description: null, + providers: ['agent-skills'], + sourceKind: 'home', + sourceLabel: 'Shared', + rootPath: SKILLS_ROOT, + directoryPath: `${SKILLS_ROOT}/${name}`, + skillFilePath: `${SKILLS_ROOT}/${name}/SKILL.md`, + installed: true, + updatedAt: null + })) + const source = { + id: 'shared', + label: 'Shared', + path: SKILLS_ROOT, + sourceKind: 'home' as const, + providers: ['agent-skills' as const], + owner: null, + exists: true + } + return { skills, sources: [source], scannedAt: 0 } +} + +it('discovers a newly installed global skill when the same chat reopens its picker', async () => { + const { result, rerender } = renderHook( + (open: boolean) => useNativeChatSkills('codex', tabId, open), + { initialProps: true } + ) + await waitFor(() => expect(result.current.status).toBe('ready')) + expect(result.current.skills).toEqual([]) + rerender(false) + installed.push('installed-later') + rerender(true) + // Reopening shows the last answer while the host is asked again, not a loading flash. + expect(result.current.status).toBe('ready') + await waitFor(() => + expect(result.current.skills.map((skill) => skill.name)).toEqual(['installed-later']) + ) +}) + +it('does not restore a closed picker snapshot when its pending scan finishes', async () => { + const release = Promise.withResolvers() + rpc.discover.mockImplementationOnce(() => release.promise) + const { result, rerender } = renderHook( + (open: boolean) => useNativeChatSkills('codex', tabId, open), + { initialProps: true } + ) + const beforeInstall = inventory() + rerender(false) + installed.push('installed-later') + await act(async () => { + release.resolve(beforeInstall) + await release.promise + }) + rerender(true) + await waitFor(() => + expect(result.current.skills.map((skill) => skill.name)).toEqual(['installed-later']) + ) +}) diff --git a/src/renderer/src/components/native-chat/composer/pickers/use-native-chat-skills.ts b/src/renderer/src/components/native-chat/composer/pickers/use-native-chat-skills.ts index ae7e2ceee82..39b910f4bfe 100644 --- a/src/renderer/src/components/native-chat/composer/pickers/use-native-chat-skills.ts +++ b/src/renderer/src/components/native-chat/composer/pickers/use-native-chat-skills.ts @@ -80,7 +80,10 @@ export function useNativeChatSkills( ) const [state, setState] = useState(IDLE_STATE) const [retryGeneration, setRetryGeneration] = useState(0) + // Last answer per pane, shown at once on reopen while the host is asked again. const paneDiscoveryCache = useRef(new Map()) + // Panes already asked during this open; store churn re-runs the effect and must not re-query. + const checkedThisOpen = useRef(new Set()) // Consume retry intent once; a generation-wide force flag would leak across pane switches. const forceNextDiscovery = useRef(false) const profile = getNativeChatAgentProfile(agent) @@ -90,6 +93,8 @@ export function useNativeChatSkills( if (!profile || !enabled || !context) { // Without a current pane, discard retry intent before it can force another pane’s scan. forceNextDiscovery.current = false + // The next open asks the host again so skills installed meanwhile appear. + checkedThisOpen.current.clear() setState(IDLE_STATE) return } @@ -111,16 +116,19 @@ export function useNativeChatSkills( const paneCacheKey = context.key const cached = paneDiscoveryCache.current.get(paneCacheKey) - if (cached) { + setState( + cached + ? { status: 'ready', skills: cached.skills, error: null, contextKey: context.key } + : { status: 'loading', skills: [], error: null, contextKey: context.key } + ) + if (cached && checkedThisOpen.current.has(paneCacheKey)) { emitNativeChatSkillDiscovery({ agent, outcome: 'ready', executionHostKind: context.executionHostKind }) - setState({ status: 'ready', skills: cached.skills, error: null, contextKey: context.key }) return } - setState({ status: 'loading', skills: [], error: null, contextKey: context.key }) // Explicit Retry bypasses shared host scans; passive discovery reuses them. const forced = forceNextDiscovery.current forceNextDiscovery.current = false @@ -128,9 +136,11 @@ export function useNativeChatSkills( void request.then( (result) => { paneDiscoveryCache.current.set(paneCacheKey, result) + // A scan that settles after close still seeds the next open but must not count as its check. if (cancelled) { return } + checkedThisOpen.current.add(paneCacheKey) emitNativeChatSkillDiscovery({ agent, outcome: 'ready', @@ -149,6 +159,11 @@ export function useNativeChatSkills( outcome: timedOut ? 'timeout' : 'error', executionHostKind: context.executionHostKind }) + checkedThisOpen.current.add(paneCacheKey) + if (cached) { + // A failed recheck keeps the last good list rather than hiding it behind an error. + return + } setState({ status: 'error', skills: [], From 04d3074e664694a7c03babf2baf78159bb49d622 Mon Sep 17 00:00:00 2001 From: Neil <4138956+nwparker@users.noreply.github.com> Date: Wed, 11 Nov 2026 19:36:21 -0800 Subject: [PATCH 12/13] fix(wsl): show Connecting instead of a stale read error while a WSL host starts (#41343) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * fix(wsl): show Connecting instead of a stale read error while a WSL host starts After Start, a WSL host can take tens of seconds to boot and update its server. The editor, file tree and source control dropped the stopped notice as soon as the host left 'stopped' and showed the error from the read that failed while it was down, telling the user to go to Settings. They now keep the notice up as 'Connecting to …', and the editor's retry waits for the WSL host and reloads the file when it answers, as it already did for SSH. * fix(wsl): skip the post-Start reload when the WSL retry already loaded the file Review follow-up: once the editor's retry reloads a file when its WSL host turns ready, Start's own reload would clear and re-read it, flashing a spinner. Start now reloads only a file that is still failed. * fix(wsl): reduced-motion spinner for Connecting; SSH targets keep their own retry signal Review follow-up: the Connecting notice uses LoadingSpinner at size-4 per the style guide, and the editor reads WSL host status only for a workspace with no SSH target. * test(editor): cover a Start reload of a still-failed file; share reload option constants * fix(wsl): reduced-motion spinner on the Start button too --- .../editor/EditorConflictReviewSurface.tsx | 12 +- .../src/components/editor/EditorContent.tsx | 12 +- .../editor/EditorEditFileSurface.tsx | 12 +- .../editor/EditorFileLoadErrorView.tsx | 13 +- .../editor/editor-panel-content-types.ts | 11 ++ .../editor/useEditorPanelContentState.ts | 14 ++- .../editor/useEditorPanelFileLoadRetry.ts | 37 +++++- ...EditorPanelFileLoadRetry.wsl-host.test.tsx | 115 ++++++++++++++++++ ...orPanelHostUnresolvedContentState.test.tsx | 51 ++++++++ .../file-explorer/panel/FileExplorer.tsx | 4 +- .../tree/FileExplorerFilesTreePane.tsx | 4 +- .../tree/FileExplorerTreeStatus.tsx | 7 +- .../source-control/listing/content-status.tsx | 7 +- .../WslHostStoppedNotice.test.tsx | 36 ++++++ .../wsl-host-stopped/WslHostStoppedNotice.tsx | 38 ++++-- .../wsl-host-stopped/use-stopped-wsl-host.ts | 31 ++++- src/renderer/src/i18n/locales/en.json | 1 + .../hosts/wsl/wsl-host-stopped-status.test.ts | 25 ++++ .../hosts/wsl/wsl-host-stopped-status.ts | 9 ++ 19 files changed, 392 insertions(+), 47 deletions(-) create mode 100644 src/renderer/src/components/editor/useEditorPanelFileLoadRetry.wsl-host.test.tsx create mode 100644 src/renderer/src/components/wsl-host-stopped/WslHostStoppedNotice.test.tsx create mode 100644 src/renderer/src/runtime/hosts/wsl/wsl-host-stopped-status.test.ts diff --git a/src/renderer/src/components/editor/EditorConflictReviewSurface.tsx b/src/renderer/src/components/editor/EditorConflictReviewSurface.tsx index 492624e8f1d..958a38d92eb 100644 --- a/src/renderer/src/components/editor/EditorConflictReviewSurface.tsx +++ b/src/renderer/src/components/editor/EditorConflictReviewSurface.tsx @@ -9,12 +9,16 @@ import type { GitStatusEntry } from '../../../../shared/git/status/git-status-ty import { ConflictBanner, ConflictPlaceholderView, ConflictReviewPanel } from './ConflictComponents' import { ImageViewer, MonacoEditor } from './editor-lazy-views' import { EditorFileLoadErrorView } from './EditorFileLoadErrorView' -import type { FileContent } from './editor-panel-content-types' +import { + HOST_STARTED, + USER_RETRY, + type FileContent, + type ReloadContentOptions +} from './editor-panel-content-types' import { useTranslate } from '@/i18n/use-translate' import type { EditorConflictNavigation } from './useEditorConflictNavigation' // Retry is a user act: it may start a stopped WSL distro. -const USER_RETRY = { wakeWslDistro: true } export function EditorConflictReviewSurface({ activeFile, @@ -43,7 +47,7 @@ export function EditorConflictReviewSurface({ ) => EditorConflictNavigation | undefined handleContentChangeForFile: (file: OpenFile, content: string) => void handleSaveForFile: (file: OpenFile, content: string) => Promise - reloadContent: (file: OpenFile, options?: { wakeWslDistro?: boolean }) => void + reloadContent: (file: OpenFile, options?: ReloadContentOptions) => void }): React.JSX.Element { const t = useTranslate() const openConflictReviewFile = useAppStore((s) => s.openConflictReviewFile) @@ -163,7 +167,7 @@ export function EditorConflictReviewSurface({ worktreeId={contentFile.worktreeId} // Borrowed views reload through their owning workspace. onRetry={() => reloadContent(recordView ?? contentFile, USER_RETRY)} - onHostStarted={() => reloadContent(recordView ?? contentFile)} + onHostStarted={() => reloadContent(recordView ?? contentFile, HOST_STARTED)} /> ) diff --git a/src/renderer/src/components/editor/EditorContent.tsx b/src/renderer/src/components/editor/EditorContent.tsx index 646f86ab287..770bcb17a70 100644 --- a/src/renderer/src/components/editor/EditorContent.tsx +++ b/src/renderer/src/components/editor/EditorContent.tsx @@ -12,7 +12,12 @@ import { EditorDiffFileSurface } from './EditorDiffFileSurface' import { EditorEditFileSurface } from './EditorEditFileSurface' import { EditorFileLoadErrorView } from './EditorFileLoadErrorView' import { MarkdownPreviewSizeGate } from './markdown-preview/render-tree/MarkdownPreviewSizeGate' -import type { FileContent } from './editor-panel-content-types' +import { + HOST_STARTED, + USER_RETRY, + type FileContent, + type ReloadContentOptions +} from './editor-panel-content-types' import { buildPdfScalePreferenceKey } from './pdf-scale-preference-storage' import { useTranslate } from '@/i18n/use-translate' import { useEditorConflictNavigation } from './useEditorConflictNavigation' @@ -20,7 +25,6 @@ import { useMarkdownDocuments } from './markdown/documents/useMarkdownDocuments' import type { MarkdownRenderState } from './markdown/render-mode/markdown-render-mode' // Retry is a user act: it may start a stopped WSL distro. -const USER_RETRY = { wakeWslDistro: true } const noopCloseMarkdownTableOfContents = (): void => {} @@ -101,7 +105,7 @@ export function EditorContent({ handleDirtyStateHint: (dirty: boolean) => void handleSave: (content: string) => Promise handleSaveForFile: (file: OpenFile, content: string) => Promise - reloadContent: (file: OpenFile, options?: { wakeWslDistro?: boolean }) => void + reloadContent: (file: OpenFile, options?: ReloadContentOptions) => void }): React.JSX.Element { const t = useTranslate() const editorViewStateKey = @@ -212,7 +216,7 @@ export function EditorContent({ code={fileContent.loadErrorCode} worktreeId={activeFile.worktreeId} onRetry={() => reloadContent(activeFile, USER_RETRY)} - onHostStarted={() => reloadContent(activeFile)} + onHostStarted={() => reloadContent(activeFile, HOST_STARTED)} /> ) } diff --git a/src/renderer/src/components/editor/EditorEditFileSurface.tsx b/src/renderer/src/components/editor/EditorEditFileSurface.tsx index b47565fd46d..a529ea780fe 100644 --- a/src/renderer/src/components/editor/EditorEditFileSurface.tsx +++ b/src/renderer/src/components/editor/EditorEditFileSurface.tsx @@ -18,7 +18,12 @@ import { import type { EditorConflictNavigation } from './useEditorConflictNavigation' import { EditorFileLoadErrorView } from './EditorFileLoadErrorView' import { RecoverableRenderErrorBoundary } from '../error-boundaries/RecoverableRenderErrorBoundary' -import type { FileContent } from './editor-panel-content-types' +import { + HOST_STARTED, + USER_RETRY, + type FileContent, + type ReloadContentOptions +} from './editor-panel-content-types' import { CustomEditorExternalChangeBanner } from './external-changes/CustomEditorExternalChangeBanner' import { PluginFileTypeSuggestion } from './PluginFileTypeSuggestion' import { BinaryFileDefaultAppButton } from './BinaryFileDefaultAppButton' @@ -31,7 +36,6 @@ import PluginCustomEditorFrame from './PluginCustomEditorFrame' import PluginBytesCustomEditorSurface from './PluginBytesCustomEditorSurface' // Retry is a user act: it may start a stopped WSL distro. -const USER_RETRY = { wakeWslDistro: true } const noopEditorContentChange = (_content: string): void => {} const noopEditorSave = async (_content: string): Promise => false @@ -109,7 +113,7 @@ export function EditorEditFileSurface({ handleContentChange: (content: string) => void handleDirtyStateHint: (dirty: boolean) => void handleSave: (content: string) => Promise - reloadContent: (file: OpenFile, options?: { wakeWslDistro?: boolean }) => void + reloadContent: (file: OpenFile, options?: ReloadContentOptions) => void }): React.JSX.Element { const t = useTranslate() if (activeFile.conflict?.kind === 'conflict-placeholder') { @@ -140,7 +144,7 @@ export function EditorEditFileSurface({ code={fileContent.loadErrorCode} worktreeId={activeFile.worktreeId} onRetry={() => reloadContent(activeFile, USER_RETRY)} - onHostStarted={() => reloadContent(activeFile)} + onHostStarted={() => reloadContent(activeFile, HOST_STARTED)} /> ) } diff --git a/src/renderer/src/components/editor/EditorFileLoadErrorView.tsx b/src/renderer/src/components/editor/EditorFileLoadErrorView.tsx index 6425ca7937f..72c6605aaa1 100644 --- a/src/renderer/src/components/editor/EditorFileLoadErrorView.tsx +++ b/src/renderer/src/components/editor/EditorFileLoadErrorView.tsx @@ -6,7 +6,7 @@ import { formatFileLoadErrorMessage } from './file-load-error-copy' import { displayIpcErrorMessage } from '@/runtime/errors/ipc-error' import { isFileTooLargeCodeMessage } from '../../../../shared/file-write/file-write-disk-precondition' import { WORKTREE_HOST_UNRESOLVED_CODE } from './editor-panel-content-types' -import { useStoppedWslHost } from '../wsl-host-stopped/use-stopped-wsl-host' +import { useWslHostReadOutage } from '../wsl-host-stopped/use-stopped-wsl-host' import { WslHostStoppedNotice } from '../wsl-host-stopped/WslHostStoppedNotice' // Localize by machine code; stored English remains readable for logs without coupling upstream state checks to UI language. @@ -39,20 +39,21 @@ export function EditorFileLoadErrorView({ }: { message: string code?: string - /** The file's workspace; a stopped WSL host there gets Start instead of this error. */ + /** The file's workspace; a stopped or still-connecting WSL host there gets its notice instead of this error. */ worktreeId?: string onRetry: () => void /** Reloads after Start without asking to wake again: the click already booted the host. */ onHostStarted?: () => void }): React.JSX.Element { const t = useTranslate() - const stoppedWslHost = useStoppedWslHost(worktreeId) - if (stoppedWslHost) { + const wslHostOutage = useWslHostReadOutage(worktreeId) + if (wslHostOutage) { return (
diff --git a/src/renderer/src/components/editor/editor-panel-content-types.ts b/src/renderer/src/components/editor/editor-panel-content-types.ts index b581ff894f8..25a97c25497 100644 --- a/src/renderer/src/components/editor/editor-panel-content-types.ts +++ b/src/renderer/src/components/editor/editor-panel-content-types.ts @@ -23,6 +23,17 @@ export const WORKTREE_HOST_UNRESOLVED_CODE = 'editor_host_workspace_unresolved' export const WORKTREE_HOST_UNRESOLVED_ERROR = "The host couldn't find this file's workspace. It may have been removed, or the host may not know about it yet. Retry, or close this tab from the tab strip." +export type ReloadContentOptions = { + wakeWslDistro?: boolean + /** Skip a file that is no longer failed, such as one already reloaded when its host answered. */ + onlyIfFailed?: boolean +} + +/** A user's Retry may boot a stopped WSL distro. */ +export const USER_RETRY: ReloadContentOptions = { wakeWslDistro: true } +/** After Start: the host is already up, and the editor's retry may have reloaded the file. */ +export const HOST_STARTED: ReloadContentOptions = { onlyIfFailed: true } + export type FileContent = { mediaUrl?: string size?: number diff --git a/src/renderer/src/components/editor/useEditorPanelContentState.ts b/src/renderer/src/components/editor/useEditorPanelContentState.ts index a9acffd7d1f..c11cf92aee5 100644 --- a/src/renderer/src/components/editor/useEditorPanelContentState.ts +++ b/src/renderer/src/components/editor/useEditorPanelContentState.ts @@ -1,7 +1,7 @@ import { useCallback, useLayoutEffect, useRef, useState } from 'react' import type { OpenFile } from '@/store/slices/editor/types/open-file' import type { useAppStore } from '@/store' -import type { DiffContent, FileContent } from './editor-panel-content-types' +import type { DiffContent, FileContent, ReloadContentOptions } from './editor-panel-content-types' import { useEditorPanelExternalContentEvents, usePruneClosedEditorContent @@ -28,7 +28,7 @@ type UseEditorPanelContentStateParams = { type UseEditorPanelContentStateResult = { fileContents: Record diffContents: Record - reloadContent: (file: OpenFile, options?: { wakeWslDistro?: boolean }) => void + reloadContent: (file: OpenFile, options?: ReloadContentOptions) => void } export function useEditorPanelContentState({ @@ -42,6 +42,7 @@ export function useEditorPanelContentState({ const [fileContents, setFileContents] = useState>({}) const [diffContents, setDiffContents] = useState>({}) const diffContentsRef = useRef(diffContents) + const fileContentsRef = useRef(fileContents) const fileLoadRetryAttemptsRef = useRef>({}) // Generation fences keep slow superseded reads from overwriting newer content. const fileReadGenerationRef = useRef>({}) @@ -63,11 +64,12 @@ export function useEditorPanelContentState({ useLayoutEffect(() => { // Why: event-driven readers must only observe state from committed renders. diffContentsRef.current = diffContents + fileContentsRef.current = fileContents openFilesRef.current = openFiles editorViewModeRef.current = editorViewMode isVisibleRef.current = isVisible activeContentFileIdRef.current = activeContentFileId - }, [activeContentFileId, diffContents, editorViewMode, isVisible, openFiles]) + }, [activeContentFileId, diffContents, editorViewMode, fileContents, isVisible, openFiles]) const invalidateFileContent = useCallback((fileIds: string[]): void => { const uniqueIds = new Set(fileIds) @@ -136,7 +138,7 @@ export function useEditorPanelContentState({ // Route explicit reloads by mode so diff banners refresh diff bodies instead of plain-file caches. const reloadContent = useCallback( - (file: OpenFile, options?: { wakeWslDistro?: boolean }): void => { + (file: OpenFile, options?: ReloadContentOptions): void => { if (file.mode === 'diff') { setDiffContents((prev) => { if (!prev[file.id]) { @@ -152,6 +154,10 @@ export function useEditorPanelContentState({ }) return } + // Why: the WSL retry may already have reloaded it when the host turned ready; a second read would flash. + if (options?.onlyIfFailed && !fileContentsRef.current[file.id]?.loadError) { + return + } delete fileLoadRetryAttemptsRef.current[file.id] setFileContents((prev) => { if (!prev[file.id]) { diff --git a/src/renderer/src/components/editor/useEditorPanelFileLoadRetry.ts b/src/renderer/src/components/editor/useEditorPanelFileLoadRetry.ts index 77198df270f..3f4b52903bb 100644 --- a/src/renderer/src/components/editor/useEditorPanelFileLoadRetry.ts +++ b/src/renderer/src/components/editor/useEditorPanelFileLoadRetry.ts @@ -1,6 +1,9 @@ import { useEffect, useRef, type Dispatch, type MutableRefObject, type SetStateAction } from 'react' import type { OpenFile } from '@/store/slices/editor/types/open-file' import { useWorktreeHostConnection } from '@/lib/workspace/ownership/worktree-host-connection-phase' +import { selectWorkspaceRuntimeEnvironmentId } from '@/lib/workspace/ownership/workspace-runtime-environment' +import { useAppStateValue } from '@/store/retained-views/app-state-view' +import type { WslHostConnectionState } from '../../../../shared/wsl-host-types' import { displayIpcErrorMessage } from '@/runtime/errors/ipc-error' import { hasRuntimeRpcErrorCode } from '../../../../shared/runtime-rpc-error-code' import { isFileTooLargeMessage } from '../../../../shared/file-write/file-write-disk-precondition' @@ -60,6 +63,25 @@ export function shouldRetryFileLoadError(message: string, code?: string): boolea ) } +/** + * A WSL host's connection in the SSH phase's terms, so a read that failed while it was down waits + * for it and retries once it answers. Null for any other host. + */ +export function wslHostRetryConnection( + environmentId: string | null, + state: WslHostConnectionState | undefined +): { targetId: string; connecting: boolean; connectedEpoch: string | null } | null { + if (!environmentId || state === undefined) { + return null + } + return { + targetId: `wsl:${environmentId}`, + connecting: state === 'connecting' || state === 'reconnecting', + // Per host, not per connection: main publishes connecting before every ready, which re-arms it. + connectedEpoch: state === 'ready' ? `wsl:${environmentId}` : null + } +} + export function useEditorPanelFileLoadRetry({ activeFile, fileContents, @@ -76,9 +98,18 @@ export function useEditorPanelFileLoadRetry({ ? fileContents[activeFileLoadRetryId]?.loadErrorCode : undefined const hostConnection = useWorktreeHostConnection(activeFile?.worktreeId ?? null) - const hostTargetId = hostConnection.targetId - const hostConnecting = hostConnection.phase === 'connecting' - const connectedHostEpoch = hostConnection.connectedEpoch + const wslEnvironmentId = useAppStateValue((state) => + selectWorkspaceRuntimeEnvironmentId(state, activeFile?.worktreeId) + ) + const wslHostState = useAppStateValue((state) => + wslEnvironmentId ? state.wslHostStatuses[wslEnvironmentId]?.state : undefined + ) + // An SSH target keeps its own signal; only a workspace with none reads its WSL host. + const wslHost = + hostConnection.phase === 'local' ? wslHostRetryConnection(wslEnvironmentId, wslHostState) : null + const hostTargetId = wslHost?.targetId ?? hostConnection.targetId + const hostConnecting = wslHost?.connecting ?? hostConnection.phase === 'connecting' + const connectedHostEpoch = wslHost ? wslHost.connectedEpoch : hostConnection.connectedEpoch const seenHostRef = useRef({ targetId: hostTargetId, connectedEpoch: connectedHostEpoch }) useEffect(() => { diff --git a/src/renderer/src/components/editor/useEditorPanelFileLoadRetry.wsl-host.test.tsx b/src/renderer/src/components/editor/useEditorPanelFileLoadRetry.wsl-host.test.tsx new file mode 100644 index 00000000000..f8a63e1a6f6 --- /dev/null +++ b/src/renderer/src/components/editor/useEditorPanelFileLoadRetry.wsl-host.test.tsx @@ -0,0 +1,115 @@ +// @vitest-environment happy-dom + +import { act, useRef, useState } from 'react' +import { createRoot, type Root } from 'react-dom/client' +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' +import type { OpenFile } from '@/store/slices/editor/types/open-file' +import type { WslHostConnectionState } from '../../../../shared/wsl-host-types' +import type { FileContent } from './editor-panel-content-types' + +const host = vi.hoisted((): { state: string | undefined } => ({ state: undefined })) + +vi.mock( + '@/lib/workspace/ownership/worktree-host-connection-phase', + () => import('./local-host-test-fixture') +) +vi.mock('@/lib/workspace/ownership/workspace-runtime-environment', () => ({ + selectWorkspaceRuntimeEnvironmentId: () => 'wsl-env' +})) +vi.mock('@/store/retained-views/app-state-view', () => ({ + useAppStateValue: (select: (state: unknown) => T): T => + select({ + wslHostStatuses: host.state + ? { 'wsl-env': { environmentId: 'wsl-env', distro: 'demo', state: host.state } } + : {} + }) +})) + +import { useEditorPanelFileLoadRetry, wslHostRetryConnection } from './useEditorPanelFileLoadRetry' + +const FILE: OpenFile = { + id: 'file-1', + filePath: '/home/orca/demo/app/main.go', + relativePath: 'main.go', + worktreeId: 'repo::/home/orca/demo/app', + language: 'go', + isDirty: false, + mode: 'edit' +} +const UNREACHABLE = 'Could not connect to the remote Orca runtime.' + +function Probe({ loadFileContent }: { loadFileContent: () => Promise }): null { + const [fileContents, setFileContents] = useState>({ + [FILE.id]: { content: '', isBinary: false, loadError: UNREACHABLE } + }) + const openFilesRef = useRef([FILE]) + const fileLoadRetryAttemptsRef = useRef>({ [FILE.id]: 3 }) + useEditorPanelFileLoadRetry({ + activeFile: FILE, + fileContents, + fileLoadRetryAttemptsRef, + loadFileContent, + openFilesRef, + setFileContents + }) + return null +} + +describe('useEditorPanelFileLoadRetry on a WSL host', () => { + let container: HTMLDivElement + let root: Root + + beforeEach(() => { + vi.useFakeTimers() + container = document.createElement('div') + root = createRoot(container) + }) + + afterEach(() => { + act(() => root.unmount()) + vi.useRealTimers() + }) + + const render = (loadFileContent: () => Promise): void => + act(() => root.render()) + + it('waits while the host connects, then reloads once it answers', async () => { + const loadFileContent = vi.fn(async () => {}) + host.state = 'connecting' + render(loadFileContent) + await act(async () => { + await vi.advanceTimersByTimeAsync(10_000) + }) + expect(loadFileContent).not.toHaveBeenCalled() + + host.state = 'ready' + render(loadFileContent) + expect(loadFileContent).toHaveBeenCalledWith( + FILE.filePath, + FILE.id, + FILE.worktreeId, + FILE.relativePath + ) + }) +}) + +describe('wslHostRetryConnection', () => { + it.each<[WslHostConnectionState, boolean, string | null]>([ + ['connecting', true, null], + ['reconnecting', true, null], + ['ready', false, 'wsl:env-1'], + ['distro-stopped', false, null], + ['failed', false, null] + ])('%s', (state, connecting, connectedEpoch) => { + expect(wslHostRetryConnection('env-1', state)).toEqual({ + targetId: 'wsl:env-1', + connecting, + connectedEpoch + }) + }) + + it('is null for a host that is not a WSL host', () => { + expect(wslHostRetryConnection(null, undefined)).toBeNull() + expect(wslHostRetryConnection('ssh-env', undefined)).toBeNull() + }) +}) diff --git a/src/renderer/src/components/editor/useEditorPanelHostUnresolvedContentState.test.tsx b/src/renderer/src/components/editor/useEditorPanelHostUnresolvedContentState.test.tsx index fdca47f2249..5941c384d8e 100644 --- a/src/renderer/src/components/editor/useEditorPanelHostUnresolvedContentState.test.tsx +++ b/src/renderer/src/components/editor/useEditorPanelHostUnresolvedContentState.test.tsx @@ -66,6 +66,7 @@ import { useEditorPanelContentState } from './useEditorPanelContentState' import { FILE_LOAD_RETRY_DELAYS_MS } from './useEditorPanelFileLoadRetry' let latestFileContents: Record = {} +let latestReload: ReturnType['reloadContent'] = () => {} const EMPTY_EDITOR_VIEW_MODE = {} const HOST_READ_LATENCY_MS = 1 @@ -80,6 +81,7 @@ function Probe({ activeFile, openFiles }: { activeFile: OpenFile; openFiles: Ope editorViewMode: EMPTY_EDITOR_VIEW_MODE }) latestFileContents = state.fileContents + latestReload = state.reloadContent return null } @@ -189,4 +191,53 @@ describe('useEditorPanelContentState — host cannot resolve a mirrored file (#2 await advanceTimers(60_000) expect(closeFile).not.toHaveBeenCalled() }) + + it('skips a reload after Start when the file already loaded, and reloads one still failed', async () => { + const activeFile = { ...makeDirtyMirroredFile(), isDirty: false } + const openFiles = [activeFile] + const store = createTestStore() + store.setState({ + openFiles, + worktreesByRepo: { + project: [ + makeWorktree({ + id: activeFile.worktreeId, + repoId: 'project', + path: '/home/user/project', + hostId: 'runtime:env-1' + }) + ] + }, + setLastKnownDiskSignature: vi.fn() + }) + mocks.getState.mockImplementation(store.getState) + mocks.readRuntimeFileContent.mockResolvedValue({ content: '# notes', isBinary: false }) + + container = document.createElement('div') + document.body.appendChild(container) + root = createRoot(container) + await act(async () => { + root?.render() + }) + await advanceTimers(HOST_READ_LATENCY_MS) + expect(latestFileContents[activeFile.id]?.content).toBe('# notes') + const readsAfterLoad = mocks.readRuntimeFileContent.mock.calls.length + + await act(async () => latestReload(activeFile, { onlyIfFailed: true })) + await advanceTimers(HOST_READ_LATENCY_MS) + expect(mocks.readRuntimeFileContent).toHaveBeenCalledTimes(readsAfterLoad) + expect(latestFileContents[activeFile.id]?.content).toBe('# notes') + + // A non-retryable failure keeps the editor's own retry out of the count. + mocks.readRuntimeFileContent.mockRejectedValueOnce(new Error('ENOENT: no such file')) + await act(async () => latestReload(activeFile)) + await advanceTimers(HOST_READ_LATENCY_MS) + expect(latestFileContents[activeFile.id]?.loadError).toContain('ENOENT') + expect(mocks.readRuntimeFileContent).toHaveBeenCalledTimes(readsAfterLoad + 1) + + await act(async () => latestReload(activeFile, { onlyIfFailed: true })) + await advanceTimers(HOST_READ_LATENCY_MS) + expect(mocks.readRuntimeFileContent).toHaveBeenCalledTimes(readsAfterLoad + 2) + expect(latestFileContents[activeFile.id]?.content).toBe('# notes') + }) }) diff --git a/src/renderer/src/components/right-sidebar/file-explorer/panel/FileExplorer.tsx b/src/renderer/src/components/right-sidebar/file-explorer/panel/FileExplorer.tsx index 2900f9ae2c3..9981fad48cf 100644 --- a/src/renderer/src/components/right-sidebar/file-explorer/panel/FileExplorer.tsx +++ b/src/renderer/src/components/right-sidebar/file-explorer/panel/FileExplorer.tsx @@ -41,7 +41,7 @@ import { useFileExplorerBackgroundMenu } from './use-file-explorer-background-me import { useFileExplorerNameFilter } from './use-file-explorer-name-filter' import { useFileExplorerTreePaneState } from '../tree/use-file-explorer-tree-pane-state' import { useWslHostUnavailableReason } from '../../../../runtime/hosts/wsl/wsl-host-availability' -import { useStoppedWslHost } from '../../../wsl-host-stopped/use-stopped-wsl-host' +import { useWslHostReadOutage } from '../../../wsl-host-stopped/use-stopped-wsl-host' import { useTranslate } from '@/i18n/use-translate' import { useUiLanguageRevision } from '@/i18n/use-ui-language-revision' @@ -57,7 +57,7 @@ function FileExplorerFiles(): React.JSX.Element { runtimeEnvironmentIdOrLocalUnlessTwin(s, activeWorktreeId) ) const wslHostUnavailable = useWslHostUnavailableReason(activeWorktreeId) - const stoppedWslHost = useStoppedWslHost(activeWorktreeId) + const stoppedWslHost = useWslHostReadOutage(activeWorktreeId) const activeWorktree = useActiveWorktree() const activeRepo = useRepoById(activeWorktree?.repoId ?? null) // Why: a folder workspace has no repo row, so only the catalog owner names its SSH host. diff --git a/src/renderer/src/components/right-sidebar/file-explorer/tree/FileExplorerFilesTreePane.tsx b/src/renderer/src/components/right-sidebar/file-explorer/tree/FileExplorerFilesTreePane.tsx index 5e247d7139c..14a84742b89 100644 --- a/src/renderer/src/components/right-sidebar/file-explorer/tree/FileExplorerFilesTreePane.tsx +++ b/src/renderer/src/components/right-sidebar/file-explorer/tree/FileExplorerFilesTreePane.tsx @@ -17,7 +17,7 @@ import type { useFileExplorerSelection } from '../selection/useFileExplorerSelec import type { useFileExplorerTree } from './useFileExplorerTree' import type { useFileExplorerTreePaneState } from './use-file-explorer-tree-pane-state' import type { WslHostUnavailableReason } from '../../../../runtime/hosts/wsl/wsl-host-availability' -import type { StoppedWslHost } from '../../../wsl-host-stopped/use-stopped-wsl-host' +import type { WslHostReadOutage } from '../../../wsl-host-stopped/use-stopped-wsl-host' type FileExplorerFilesTreePaneProps = { activeRepo: Repo | null @@ -39,7 +39,7 @@ type FileExplorerFilesTreePaneProps = { handleExplorerBackgroundContextMenuCapture: (event: React.MouseEvent) => void handleExplorerBackgroundDoubleClick: (event: React.MouseEvent) => void wslHostUnavailable?: WslHostUnavailableReason | null - stoppedWslHost?: StoppedWslHost | null + stoppedWslHost?: WslHostReadOutage | null } export function FileExplorerFilesTreePane({ diff --git a/src/renderer/src/components/right-sidebar/file-explorer/tree/FileExplorerTreeStatus.tsx b/src/renderer/src/components/right-sidebar/file-explorer/tree/FileExplorerTreeStatus.tsx index 12cbcf2c5ae..69594ffcb3a 100644 --- a/src/renderer/src/components/right-sidebar/file-explorer/tree/FileExplorerTreeStatus.tsx +++ b/src/renderer/src/components/right-sidebar/file-explorer/tree/FileExplorerTreeStatus.tsx @@ -7,7 +7,7 @@ import { getWslHostsTurnedOffLabel, getWslHostsTurnOnLabel } from '../../../sidebar/wsl-host-row-status' -import type { StoppedWslHost } from '../../../wsl-host-stopped/use-stopped-wsl-host' +import type { WslHostReadOutage } from '../../../wsl-host-stopped/use-stopped-wsl-host' import { WslHostStoppedNotice } from '../../../wsl-host-stopped/WslHostStoppedNotice' type FileExplorerTreeStatusProps = { @@ -18,8 +18,8 @@ type FileExplorerTreeStatusProps = { emptyMessage?: string /** The workspace's WSL host is turned off: offer the switch instead of a read error. */ wslHostsTurnedOff?: boolean - /** The workspace's WSL host is stopped: offer Start instead of a read error. */ - stoppedWslHost?: StoppedWslHost | null + /** The workspace's WSL host is stopped or still connecting: its notice instead of a read error. */ + stoppedWslHost?: WslHostReadOutage | null onWslHostStarted?: () => void } @@ -62,6 +62,7 @@ export function FileExplorerTreeStatus({ diff --git a/src/renderer/src/components/right-sidebar/source-control/listing/content-status.tsx b/src/renderer/src/components/right-sidebar/source-control/listing/content-status.tsx index 29efa174c67..3cec349ba3f 100644 --- a/src/renderer/src/components/right-sidebar/source-control/listing/content-status.tsx +++ b/src/renderer/src/components/right-sidebar/source-control/listing/content-status.tsx @@ -7,7 +7,7 @@ import { ConflictSummaryCard, OperationBanner } from './conflict-status-cards' import { EmptyState } from './empty-state' import { StatusRetryButton } from './status-retry-button' import { TooManyChangesBanner } from './too-many-changes-banner' -import { useStoppedWslHost } from '../../../wsl-host-stopped/use-stopped-wsl-host' +import { useWslHostReadOutage } from '../../../wsl-host-stopped/use-stopped-wsl-host' import { WslHostStoppedNotice } from '../../../wsl-host-stopped/WslHostStoppedNotice' import { useSourceControlScope } from '../panel/worktree-scope' import { shortRefLabel } from '../../../../../../shared/source-control/short-ref-label' @@ -62,8 +62,8 @@ export function SourceControlContentStatus({ filterQuery: string }): React.JSX.Element { const t = useTranslate() - // Why: a stopped host can't answer git, so its read error or "clean" result is not the workspace's. - const stoppedWslHost = useStoppedWslHost(worktreeId) + // Why: a stopped or still-connecting host can't answer git, so its read error or "clean" result is not the workspace's. + const stoppedWslHost = useWslHostReadOutage(worktreeId) // Folder sections keep to the plain message and Retry; raw Git output stays in the worktree's own panel. const showRawStatusError = useSourceControlScope() === null return ( @@ -72,6 +72,7 @@ export function SourceControlContentStatus({ )} diff --git a/src/renderer/src/components/wsl-host-stopped/WslHostStoppedNotice.test.tsx b/src/renderer/src/components/wsl-host-stopped/WslHostStoppedNotice.test.tsx new file mode 100644 index 00000000000..5e1c02f613b --- /dev/null +++ b/src/renderer/src/components/wsl-host-stopped/WslHostStoppedNotice.test.tsx @@ -0,0 +1,36 @@ +// @vitest-environment happy-dom + +import { act } from 'react' +import { createRoot, type Root } from 'react-dom/client' +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' + +vi.mock('@/runtime/hosts/wsl/wsl-host-user-start', () => ({ + startWslHostAndReloadProjects: vi.fn(async () => true) +})) + +import { WslHostStoppedNotice } from './WslHostStoppedNotice' + +describe('WslHostStoppedNotice', () => { + let container: HTMLDivElement + let root: Root + + beforeEach(() => { + container = document.createElement('div') + root = createRoot(container) + }) + + afterEach(() => act(() => root.unmount())) + + it('offers Start while the host is stopped', () => { + act(() => root.render()) + expect(container.textContent).toContain('demo is stopped') + expect(container.querySelector('button')?.textContent).toBe('Start') + }) + + it('shows Connecting with no Start while the host boots', () => { + act(() => root.render()) + expect(container.textContent).toContain('Connecting to demo…') + expect(container.querySelector('[data-wsl-host-connecting]')).not.toBeNull() + expect(container.querySelector('button')).toBeNull() + }) +}) diff --git a/src/renderer/src/components/wsl-host-stopped/WslHostStoppedNotice.tsx b/src/renderer/src/components/wsl-host-stopped/WslHostStoppedNotice.tsx index 1d8a7d3589f..f9abb57d667 100644 --- a/src/renderer/src/components/wsl-host-stopped/WslHostStoppedNotice.tsx +++ b/src/renderer/src/components/wsl-host-stopped/WslHostStoppedNotice.tsx @@ -1,22 +1,27 @@ import React, { useState } from 'react' -import { Loader2 } from 'lucide-react' import { Button } from '@/components/ui/button' +import { LoadingSpinner } from '@/components/LoadingSpinner' import { LinuxPenguinIcon } from '@/components/icons/LinuxPenguinIcon' import { useMountedRef } from '@/hooks/useMountedRef' import { useTranslate } from '@/i18n/use-translate' import { cn } from '@/lib/class-names' import { startWslHostAndReloadProjects } from '@/runtime/hosts/wsl/wsl-host-user-start' -/** " is stopped" with Start, where a surface would otherwise show a read error. Only the click boots (E9). */ +/** + * " is stopped" with Start, where a surface would otherwise show a read error. Only the click + * boots (E9). `connecting` swaps in "Connecting to …" while it boots, so the stale error never shows. + */ export function WslHostStoppedNotice({ environmentId, distro, + connecting = false, size = 'xs', className, onStarted }: { environmentId: string distro: string + connecting?: boolean size?: 'xs' | 'sm' className?: string onStarted?: () => void @@ -48,15 +53,26 @@ export function WslHostStoppedNotice({ className="mb-1 size-7 text-muted-foreground" data-wsl-host-stopped-icon="" /> - - {t('components.native-chat.hostOutage.stopped', '{{distro}} is stopped', { - distro - })} - - + {connecting ? ( + + + {t('components.native-chat.hostOutage.connecting', 'Connecting to {{distro}}…', { + distro + })} + + ) : ( + <> + + {t('components.native-chat.hostOutage.stopped', '{{distro}} is stopped', { + distro + })} + + + + )}
) } diff --git a/src/renderer/src/components/wsl-host-stopped/use-stopped-wsl-host.ts b/src/renderer/src/components/wsl-host-stopped/use-stopped-wsl-host.ts index 320b45092a9..8f510d1b0c5 100644 --- a/src/renderer/src/components/wsl-host-stopped/use-stopped-wsl-host.ts +++ b/src/renderer/src/components/wsl-host-stopped/use-stopped-wsl-host.ts @@ -1,11 +1,16 @@ import { useAppStateValue } from '@/store/retained-views/app-state-view' import { useMemo } from 'react' import type { AppState } from '@/store/types' -import { selectStoppedWslDistro } from '@/runtime/hosts/wsl/wsl-host-stopped-status' +import { + selectConnectingWslDistro, + selectStoppedWslDistro +} from '@/runtime/hosts/wsl/wsl-host-stopped-status' import { selectWorkspaceRuntimeEnvironmentId } from '../../lib/workspace/ownership/workspace-runtime-environment' import { selectWslHostEnvironmentUnavailableReason } from '../../runtime/hosts/wsl/wsl-host-availability' export type StoppedWslHost = { environmentId: string; distro: string } +/** A host a read surface waits on instead of showing its read error; `connecting` once Start (or a reconnect) is under way. */ +export type WslHostReadOutage = StoppedWslHost & { connecting: boolean } /** A stopped distro Start can boot; a host that cannot run (turned off, server files missing) keeps its own notice. */ export function selectStartableStoppedWslDistro( @@ -40,3 +45,27 @@ export function useStoppedWslHost(worktreeId: string | null | undefined): Stoppe [environmentId, distro] ) } + +/** + * For surfaces that mask a read error: a read that failed while the host was down stays stale + * until the host answers, so a connecting host keeps the notice up instead of that error. + */ +export function useWslHostReadOutage( + worktreeId: string | null | undefined +): WslHostReadOutage | null { + const stopped = useStoppedWslHost(worktreeId) + const environmentId = useAppStateValue((state) => + selectWorkspaceRuntimeEnvironmentId(state, worktreeId) + ) + const connectingDistro = useAppStateValue((state) => + selectConnectingWslDistro(state, environmentId) + ) + return useMemo(() => { + if (stopped) { + return { ...stopped, connecting: false } + } + return environmentId && connectingDistro + ? { environmentId, distro: connectingDistro, connecting: true } + : null + }, [stopped, environmentId, connectingDistro]) +} diff --git a/src/renderer/src/i18n/locales/en.json b/src/renderer/src/i18n/locales/en.json index 48e2502cb50..2abec80a0f5 100644 --- a/src/renderer/src/i18n/locales/en.json +++ b/src/renderer/src/i18n/locales/en.json @@ -21169,6 +21169,7 @@ "reconnect": "Reconnect", "stopped": "{{distro}} is stopped", "start": "Start", + "connecting": "Connecting to {{distro}}…", "startFailed": "Couldn’t start {{distro}}. Check it in Settings → Remote Orca Servers.", "unverifiable": "Orca couldn't check whether {{distro}} is running", "startToSend": "Start {{distro}} to send.", diff --git a/src/renderer/src/runtime/hosts/wsl/wsl-host-stopped-status.test.ts b/src/renderer/src/runtime/hosts/wsl/wsl-host-stopped-status.test.ts new file mode 100644 index 00000000000..f7f8c70ecfb --- /dev/null +++ b/src/renderer/src/runtime/hosts/wsl/wsl-host-stopped-status.test.ts @@ -0,0 +1,25 @@ +import { describe, expect, it } from 'vitest' +import type { WslHostConnectionState } from '../../../../../shared/wsl-host-types' +import { selectConnectingWslDistro, selectStoppedWslDistro } from './wsl-host-stopped-status' + +const stateWith = (state: WslHostConnectionState) => ({ + wslHostStatuses: { env: { environmentId: 'env', distro: 'demo', state, detail: null } } +}) + +describe('selectConnectingWslDistro', () => { + it.each<[WslHostConnectionState, string | null]>([ + ['connecting', 'demo'], + ['reconnecting', 'demo'], + ['ready', null], + ['distro-stopped', null], + ['failed', null] + ])('%s', (state, expected) => { + expect(selectConnectingWslDistro(stateWith(state), 'env')).toBe(expected) + }) + + it('is null for an unknown host and never overlaps stopped', () => { + expect(selectConnectingWslDistro(stateWith('connecting'), 'other')).toBeNull() + expect(selectConnectingWslDistro(stateWith('connecting'), null)).toBeNull() + expect(selectStoppedWslDistro(stateWith('connecting'), 'env')).toBeNull() + }) +}) diff --git a/src/renderer/src/runtime/hosts/wsl/wsl-host-stopped-status.ts b/src/renderer/src/runtime/hosts/wsl/wsl-host-stopped-status.ts index 6e814a8ee1e..892431c1a92 100644 --- a/src/renderer/src/runtime/hosts/wsl/wsl-host-stopped-status.ts +++ b/src/renderer/src/runtime/hosts/wsl/wsl-host-stopped-status.ts @@ -9,6 +9,15 @@ export function selectStoppedWslDistro( return status?.state === 'distro-stopped' ? status.distro : null } +/** The distro of a WSL host still connecting, such as after Start while it boots or updates its server. */ +export function selectConnectingWslDistro( + state: Pick, + environmentId: string | null | undefined +): string | null { + const status = environmentId ? state.wslHostStatuses[environmentId] : undefined + return status?.state === 'connecting' || status?.state === 'reconnecting' ? status.distro : null +} + /** * Whether a pane shows the stopped distro's Start card. A pane whose create failed has no server * yet and is still the workspace's; one on another server, or a host that cannot run, keeps its own From 3f84308ec11b8068aaec4b637ed1ced99f14213c Mon Sep 17 00:00:00 2001 From: Brennan Benson <79079362+brennanb2025@users.noreply.github.com> Date: Wed, 11 Nov 2026 19:36:32 -0800 Subject: [PATCH 13/13] Revert orca chat list/read/send/wait (#40836) and its perf repair (#41129) (#41241) * Revert "Make chat wait check the chat's work state instead of reading its history every second (#41129)" This reverts commit 9a10464bec869959fe658640b62800c51878ea89. * Revert "feat(cli): read, message and wait on agent chats with orca chat list/read/send/wait (#40836)" This reverts commit c2e6ebbf0b67e93241d58db6290d2292e95f5fdc. --- config/tsconfig.node.json | 3 +- config/tsconfig.test.pane-launch.json | 1 - config/vitest.performance.config.ts | 3 +- docs/site/content/docs/cli/reference.mdx | 15 -- skill-guides/orca-cli.md | 30 +-- src/cli/bundled-skill-guides.ts | 30 +-- src/cli/chat-history.ts | 206 ------------------ src/cli/chat-sessions.ts | 175 --------------- src/cli/command-scoped-flag-help.ts | 15 -- src/cli/flag-help-text.ts | 1 - src/cli/handler-group-manifest.ts | 2 +- src/cli/handlers/chat.test.ts | 159 -------------- src/cli/handlers/chat.ts | 191 +++------------- src/cli/root-help-text-primary.ts | 4 - src/cli/specs/chat.ts | 52 ----- .../input-order/journal-newest-turn.ts | 7 - .../agent-session-journal/journal-store.ts | 6 +- .../host/structured-agent-session-host.ts | 3 +- ...d-agent-session-background-task-channel.ts | 28 --- ...agent-session-chat-wait-work-state.test.ts | 124 ----------- .../structured-agent-session.ts | 11 - ...ructured-codex-session-rpc-test-harness.ts | 10 +- .../agent-session/agent-session-work-state.ts | 12 - .../rpc-params-catalog.generated.ts | 4 +- .../structured-agent-session-params.ts | 2 - 25 files changed, 42 insertions(+), 1052 deletions(-) delete mode 100644 src/cli/chat-history.ts delete mode 100644 src/cli/chat-sessions.ts delete mode 100644 src/main/runtime/agent-session/structured-agent-session-chat-wait-work-state.test.ts delete mode 100644 src/shared/agent-session/agent-session-work-state.ts diff --git a/config/tsconfig.node.json b/config/tsconfig.node.json index c616a385638..86dad6f7d38 100644 --- a/config/tsconfig.node.json +++ b/config/tsconfig.node.json @@ -39,8 +39,7 @@ "../src/types/**/*" ], "exclude": [ - "../src/main/runtime/orchestration/terminal-authority/structured-sessions/structured-chat-coordinator-mail.test.ts", - "../src/main/runtime/agent-session/structured-agent-session-chat-wait-work-state.test.ts" + "../src/main/runtime/orchestration/terminal-authority/structured-sessions/structured-chat-coordinator-mail.test.ts" ], "compilerOptions": { "composite": true, diff --git a/config/tsconfig.test.pane-launch.json b/config/tsconfig.test.pane-launch.json index 07117e2ab40..cdffd47c2df 100644 --- a/config/tsconfig.test.pane-launch.json +++ b/config/tsconfig.test.pane-launch.json @@ -2,7 +2,6 @@ "extends": "./tsconfig.web.json", "include": [ "../src/main/runtime/orchestration/terminal-authority/structured-sessions/structured-chat-coordinator-mail.test.ts", - "../src/main/runtime/agent-session/structured-agent-session-chat-wait-work-state.test.ts", "../src/renderer/src/components/task-page/linear/task-page-linear-detail-query.test.tsx", "../src/renderer/src/runtime/observed-provider-accounts.test.tsx", "../src/renderer/src/components/status-bar/ClaudeSwitcherMenu.observation.test.tsx", diff --git a/config/vitest.performance.config.ts b/config/vitest.performance.config.ts index 07688401e26..37d4f7a4763 100644 --- a/config/vitest.performance.config.ts +++ b/config/vitest.performance.config.ts @@ -14,8 +14,7 @@ const contracts = [ 'src/main/codex-usage/scanner-incremental-append.test.ts', 'src/main/usage/source-input/jsonl/jsonl-file-reader.test.ts', 'config/scripts/static-analysis/oxlint-contracts/app-store-performance-plugin.test.mjs', - 'config/scripts/static-analysis/oxlint-contracts/quadratic-buffer-concat-plugin.test.mjs', - 'src/main/runtime/agent-session/structured-agent-session-chat-wait-work-state.test.ts' + 'config/scripts/static-analysis/oxlint-contracts/quadratic-buffer-concat-plugin.test.mjs' ] for (const contract of contracts) { if (!existsSync(resolve(contract))) { diff --git a/docs/site/content/docs/cli/reference.mdx b/docs/site/content/docs/cli/reference.mdx index bff53fb760b..ac3bfc1ba6c 100644 --- a/docs/site/content/docs/cli/reference.mdx +++ b/docs/site/content/docs/cli/reference.mdx @@ -233,21 +233,6 @@ orca chat rename --name "Migrate auth" Renames a chat tab, the same as renaming it by hand. `--session` takes the `sessionId` that `orca agent launch` prints. Run inside a chat, it defaults to that chat; with `--environment` it is required. -Agents that run as chats are read and messaged with `chat` commands. `--session` takes the session id from `chat list` or the handle `agent launch` returns. - -```bash -orca chat list --json -orca chat list --state working --json -orca chat read --session --limit 10 --json -orca chat send --session --text "Run the tests, then summarize failures" --json -orca chat send --session --text "" --wait --timeout-ms 600000 --json -orca chat wait --session --timeout-ms 600000 --json -``` - -`chat read` returns the newest messages, tool calls, errors, approvals and questions; an approval or question waiting for an answer has `pending: true`. `chat send` sends the text as your message. While the agent is working, hosts with a message queue hold it until the turn ends and report `delivery: "queued"`. A chat that restarted since it was read refuses the send. `--text-file -` reads the message from stdin. - -`chat send --wait` and `chat wait` return once the chat's own turn settles, in the words `chat list` uses: `settled: "done"`, or `settled: "blocked"` when an approval or question needs an answer. Background agents the chat started can still show as `working` or `waiting` in `chat list`. They print the entries since the last user message. A wait past `--timeout-ms` returns `settled: "timeout"`, and a message held in a stopped or failed queue returns `settled: "held"`; both exit 1. `chat read --json` returns a `cursor` for the newest row read, and `truncated: true` when older rows it needed were past the 200-row read. `chat list --state` takes `working`, `blocked`, `waiting` or `done`. Paired hosts (`--environment`, `--pairing-code`) can list chats but not read, send or wait yet. - ## Files ```bash diff --git a/skill-guides/orca-cli.md b/skill-guides/orca-cli.md index ddd4193d06f..62f9ce9abb1 100644 --- a/skill-guides/orca-cli.md +++ b/skill-guides/orca-cli.md @@ -28,9 +28,9 @@ Prefer `--json` for agent-driven calls. If the CLI is missing, say so explicitly A full handoff transfers ownership to another agent or worktree, then the original agent stops. Treat requests phrased as "hand off", "handoff", "handover", "give this to another agent", "give this to another worktree", "another agent", or "another worktree" as full handoffs unless the user expects the result back ("have another agent review this and report") or explicitly asks to supervise, monitor, wait for results, track completion, coordinate a DAG, use decision gates, or manage ask/reply. -A handoff is done when the new worktree id and agent handle have been reported and the prompt was delivered: the launch's `prompt.outcome` is `journaled` or `handed-to-terminal`, or, for an agent that was already running, the `terminal send` receipt reported `accepted: true` or `chat send` reported `delivery` `submitted` or `queued`. Do not wait for the receiving agent to finish. +A handoff is done when the new worktree id and agent handle have been reported and the prompt was delivered: the launch's `prompt.outcome` is `journaled` or `handed-to-terminal`, or, for an agent that was already running, the send receipt reported `accepted: true`. Do not wait for the receiving agent to finish. -Do not use `orca orchestration task-create`, `orca orchestration dispatch --inject`, or `orca orchestration check --wait` for full handoffs. `task-create` is also forbidden because it records coordinator-owned tracking state; if a task row is needed, the user asked for supervised orchestration. Deliver the prompt with the agent launch, or with `terminal send` or `chat send` for an agent that is already running. +Do not use `orca orchestration task-create`, `orca orchestration dispatch --inject`, or `orca orchestration check --wait` for full handoffs. `task-create` is also forbidden because it records coordinator-owned tracking state; if a task row is needed, the user asked for supervised orchestration. Deliver the prompt with the agent launch, or with terminal commands for an agent that is already running. For local work in the caller’s project, omit `--repo` and `--project`. Pass one to target another project or when Orca asks you to disambiguate; do not pick a repo from a list for unambiguous same-project work. The project comes from the caller’s directory, not the project visible in the app. Paired-server calls require an explicit server-side target. @@ -62,14 +62,6 @@ Existing-terminal handoff: ORCA terminal send --terminal --text "" --enter --json ``` -Existing-chat handoff, for an agent that runs as a chat (`outcome.kind` `structured`): - -```text -ORCA chat send --session --text "" --json -ORCA chat send --session --text "" --wait --timeout-ms 600000 --json -ORCA chat wait --session --timeout-ms 600000 --json -``` - ## Starting agents `worktree create --agent` (new checkout) and `agent launch` (existing worktree) are how to start an agent. They go through Orca's agent launch, the same one the app's new-agent tab uses: the user's Chat | CLI choice for that agent decides whether it opens as a chat or a terminal, and the result reports which (`outcome.kind`) and why (`receipt.detail`). When a chat is not possible, such as for an agent without chat support, the launch starts a terminal and the receipt says so. `--surface chat|terminal` picks one for that launch instead of the setting; pass it only when the user asked for a chat or a terminal. `--surface chat` is refused when Orca can establish before starting anything that the agent cannot open as a chat. If support cannot be checked, the launch can fall back to a terminal. If chat becomes unavailable only after a new workspace is created, Orca keeps that workspace, its setup and default tabs, then opens a terminal; the receipt explains the fallback. `terminal create` makes a terminal for shells and commands. @@ -281,24 +273,6 @@ Terminal rules: - For long output, use cursor reads. After a limited tail preview, page from `oldestCursor`; after a cursor read, continue with `nextCursor` while `limited` is true and `nextCursor !== latestCursor`. - `--direction vertical` splits left/right. `--direction horizontal` splits top/bottom. -## Chats - -Agents launched as chats are read and messaged with `chat`, not `terminal`. `chat list` names each chat's workspace, agent, state and last reply; `--session` takes the `session` it reports or the `handle` a launch returned. - -```text -ORCA chat list --json -ORCA chat list --state working --json -ORCA chat read --session --limit 10 --json -ORCA chat send --session --text "" --json -``` - -- `chat read` returns messages, tool calls, errors, and approvals or questions; `pending: true` marks one waiting for the user. -- `chat send` sends the text as the user's message. While the agent works, a host with a message queue holds it until the turn ends and reports `delivery: "queued"`. -- A refused send was not delivered. Read the chat again before resending. -- `chat send --wait` (or `chat wait`) returns when the chat settles, using `chat list` states: `settled: "done"`, or `"blocked"` on an approval or question, with the entries since the last user message; `settled: "timeout"` or `"held"` (queued message stopped or failed) exits 1. Pass `--timeout-ms`. Long or quoted text goes through `--text-file -` on stdin. -- Paired hosts can `chat list` only; read, send and wait run on the host machine. -- When relaying for a user from outside Orca (for example a voice or messaging assistant on the same machine), report from `chat list` and `chat read`, repeat an instruction back and get the user's confirmation before `chat send`, and send the user's words without adding instructions. - ## Artifacts diff --git a/src/cli/bundled-skill-guides.ts b/src/cli/bundled-skill-guides.ts index d8f59db7cdf..166823d580b 100644 --- a/src/cli/bundled-skill-guides.ts +++ b/src/cli/bundled-skill-guides.ts @@ -256,9 +256,9 @@ const ORCA_CLI_MARKDOWN = [ "", "A full handoff transfers ownership to another agent or worktree, then the original agent stops. Treat requests phrased as \"hand off\", \"handoff\", \"handover\", \"give this to another agent\", \"give this to another worktree\", \"another agent\", or \"another worktree\" as full handoffs unless the user expects the result back (\"have another agent review this and report\") or explicitly asks to supervise, monitor, wait for results, track completion, coordinate a DAG, use decision gates, or manage ask/reply.", "", - "A handoff is done when the new worktree id and agent handle have been reported and the prompt was delivered: the launch's `prompt.outcome` is `journaled` or `handed-to-terminal`, or, for an agent that was already running, the `terminal send` receipt reported `accepted: true` or `chat send` reported `delivery` `submitted` or `queued`. Do not wait for the receiving agent to finish.", + "A handoff is done when the new worktree id and agent handle have been reported and the prompt was delivered: the launch's `prompt.outcome` is `journaled` or `handed-to-terminal`, or, for an agent that was already running, the send receipt reported `accepted: true`. Do not wait for the receiving agent to finish.", "", - "Do not use `orca orchestration task-create`, `orca orchestration dispatch --inject`, or `orca orchestration check --wait` for full handoffs. `task-create` is also forbidden because it records coordinator-owned tracking state; if a task row is needed, the user asked for supervised orchestration. Deliver the prompt with the agent launch, or with `terminal send` or `chat send` for an agent that is already running.", + "Do not use `orca orchestration task-create`, `orca orchestration dispatch --inject`, or `orca orchestration check --wait` for full handoffs. `task-create` is also forbidden because it records coordinator-owned tracking state; if a task row is needed, the user asked for supervised orchestration. Deliver the prompt with the agent launch, or with terminal commands for an agent that is already running.", "", "For local work in the caller’s project, omit `--repo` and `--project`. Pass one to target another project or when Orca asks you to disambiguate; do not pick a repo from a list for unambiguous same-project work. The project comes from the caller’s directory, not the project visible in the app. Paired-server calls require an explicit server-side target.", "", @@ -290,14 +290,6 @@ const ORCA_CLI_MARKDOWN = [ "ORCA terminal send --terminal --text \"\" --enter --json", "```", "", - "Existing-chat handoff, for an agent that runs as a chat (`outcome.kind` `structured`):", - "", - "```text", - "ORCA chat send --session --text \"\" --json", - "ORCA chat send --session --text \"\" --wait --timeout-ms 600000 --json", - "ORCA chat wait --session --timeout-ms 600000 --json", - "```", - "", "## Starting agents", "", "`worktree create --agent` (new checkout) and `agent launch` (existing worktree) are how to start an agent. They go through Orca's agent launch, the same one the app's new-agent tab uses: the user's Chat | CLI choice for that agent decides whether it opens as a chat or a terminal, and the result reports which (`outcome.kind`) and why (`receipt.detail`). When a chat is not possible, such as for an agent without chat support, the launch starts a terminal and the receipt says so. `--surface chat|terminal` picks one for that launch instead of the setting; pass it only when the user asked for a chat or a terminal. `--surface chat` is refused when Orca can establish before starting anything that the agent cannot open as a chat. If support cannot be checked, the launch can fall back to a terminal. If chat becomes unavailable only after a new workspace is created, Orca keeps that workspace, its setup and default tabs, then opens a terminal; the receipt explains the fallback. `terminal create` makes a terminal for shells and commands.", @@ -509,24 +501,6 @@ const ORCA_CLI_MARKDOWN = [ "- For long output, use cursor reads. After a limited tail preview, page from `oldestCursor`; after a cursor read, continue with `nextCursor` while `limited` is true and `nextCursor !== latestCursor`.", "- `--direction vertical` splits left/right. `--direction horizontal` splits top/bottom.", "", - "## Chats", - "", - "Agents launched as chats are read and messaged with `chat`, not `terminal`. `chat list` names each chat's workspace, agent, state and last reply; `--session` takes the `session` it reports or the `handle` a launch returned.", - "", - "```text", - "ORCA chat list --json", - "ORCA chat list --state working --json", - "ORCA chat read --session --limit 10 --json", - "ORCA chat send --session --text \"\" --json", - "```", - "", - "- `chat read` returns messages, tool calls, errors, and approvals or questions; `pending: true` marks one waiting for the user.", - "- `chat send` sends the text as the user's message. While the agent works, a host with a message queue holds it until the turn ends and reports `delivery: \"queued\"`.", - "- A refused send was not delivered. Read the chat again before resending.", - "- `chat send --wait` (or `chat wait`) returns when the chat settles, using `chat list` states: `settled: \"done\"`, or `\"blocked\"` on an approval or question, with the entries since the last user message; `settled: \"timeout\"` or `\"held\"` (queued message stopped or failed) exits 1. Pass `--timeout-ms`. Long or quoted text goes through `--text-file -` on stdin.", - "- Paired hosts can `chat list` only; read, send and wait run on the host machine.", - "- When relaying for a user from outside Orca (for example a voice or messaging assistant on the same machine), report from `chat list` and `chat read`, repeat an instruction back and get the user's confirmation before `chat send`, and send the user's words without adding instructions.", - "", "", "", "## Artifacts", diff --git a/src/cli/chat-history.ts b/src/cli/chat-history.ts deleted file mode 100644 index aad6b96ddeb..00000000000 --- a/src/cli/chat-history.ts +++ /dev/null @@ -1,206 +0,0 @@ -import type { - AgentSessionHistoryPage, - AgentSessionHistoryResult -} from '../shared/agent-session/agent-session-wire' -import type { AgentSessionWorkState } from '../shared/agent-session/agent-session-work-state' -import { AGENT_SESSION_HISTORY_MAX_LIMIT } from '../shared/agent-session/agent-session-wire' -import { - AGENT_SESSION_UNATTACHED_REFUSAL_CODE, - isUnattachedAgentSessionReadRefusal -} from '../shared/agent-session/failures/structured-agent-session-read-refusal' -import { structuredAgentSessionQueueSendsNext } from '../shared/agent-session/view/structured-agent-session-live-turn' -import { chatTranscript, formatChatRead, type ChatTranscriptEntry } from './chat-sessions' -import type { RuntimeClient, RuntimeRpcSuccess } from './runtime-client' -import { RuntimeClientError } from './runtime-client' - -export const DEFAULT_READ_LIMIT = 20 -const WAIT_POLL_MS = 1000 - -export type ChatTranscript = { - session: string - entries: ChatTranscriptEntry[] - /** Newest journal row read, so a later `--since` can continue from it. */ - cursor: AgentSessionHistoryPage['window']['newest'] - /** Older rows exist that this 200-row read did not reach, and they were needed. */ - truncated: boolean -} -/** The chat's own turn in `chat list` words (structuredAgentSessionLeadState), plus two wait-only outcomes. */ -export type ChatSettled = 'done' | 'blocked' | 'held' | 'timeout' - -export async function readHistory( - client: RuntimeClient, - session: string, - limit: number -): Promise> { - return client - .call('agentSession.history', { - sessionId: session, - direction: 'tail', - limit - }) - .catch((error: unknown) => { - throw explainChatRefusal(error, session, 'read') - }) -} - -export async function readTranscript( - client: RuntimeClient, - session: string, - limit?: number -): Promise> { - // Status frames and turn markers share the journal, so read the most rows a page allows. - const response = await readHistory(client, session, AGENT_SESSION_HISTORY_MAX_LIMIT) - const { page } = response.result - const all = chatTranscript(page.items, page.actionablePromptIds) - const entries = limit === undefined ? all : all.slice(-limit) - const truncated = page.hasOlder && (limit === undefined || all.length < limit) - return { ...response, result: { session, entries, cursor: page.window.newest, truncated } } -} - -/** Older hosts' history pages may omit `working`. */ -export type ChatWorkState = Omit & { working?: boolean } - -/** One command's reads of the facts a send or wait decides from. Hosts without - * `agentSession.workState` answer from a history page for the rest of the command. */ -export function chatWorkStateReader( - client: RuntimeClient, - session: string -): () => Promise { - let historyOnly = false - return async () => { - if (!historyOnly) { - try { - const response = await client.call('agentSession.workState', { - sessionId: session - }) - return response.result - } catch (error) { - if (!(error instanceof RuntimeClientError && error.code === 'method_not_found')) { - throw explainChatRefusal(error, session, 'read') - } - historyOnly = true - } - } - const { page } = (await readHistory(client, session, 1)).result - return { - ...(page.working !== undefined ? { working: page.working } : {}), - actionablePrompt: (page.actionablePromptIds?.length ?? 0) > 0, - ...(page.queuedMessages !== undefined - ? { queuedMessageIds: (page.queuedMessages ?? []).map((m) => m.messageId) } - : {}), - nextQueuedMessageId: page.nextQueuedMessageId ?? null, - ...(page.fence !== undefined ? { fence: page.fence } : {}) - } - } -} - -export function requireWorkingFlag(state: { working?: boolean }): void { - if (state.working === undefined) { - throw new RuntimeClientError( - 'incompatible_runtime', - 'This Orca host does not report when a chat is working, so it cannot wait. Update Orca on the host.' - ) - } -} - -// Settled means no running turn or unanswered send, or a question/approval waiting for the user. -export async function waitForChat( - readWorkState: () => Promise, - timeoutMs: number | undefined, - queuedMessageId?: string -): Promise { - const deadline = timeoutMs === undefined ? Infinity : Date.now() + timeoutMs - for (;;) { - const state = await readWorkState() - requireWorkingFlag(state) - if (state.actionablePrompt) { - return 'blocked' - } - // Between turns the host names its next automatic send while working is briefly false. - const sendsNext = structuredAgentSessionQueueSendsNext( - state.nextQueuedMessageId, - state.working === true - ) - if (!state.working && !sendsNext) { - // Still queued with nothing running or about to send: stopped, failed or held across a reopen. - const held = - queuedMessageId !== undefined && state.queuedMessageIds?.includes(queuedMessageId) === true - return held ? 'held' : 'done' - } - if (Date.now() >= deadline) { - return 'timeout' - } - await new Promise((resolve) => setTimeout(resolve, WAIT_POLL_MS)) - } -} - -// After a wait, the turn's entries: everything after the last user message. -export async function readSettledTurn( - client: RuntimeClient, - session: string, - settled: ChatSettled -): Promise> { - if (settled === 'timeout' || settled === 'held') { - // Chained commands such as `send --wait && ...` need a failing exit when the chat did not settle. - process.exitCode = 1 - } - const response = await readTranscript(client, session) - const { entries, truncated } = response.result - const lastUser = entries.findLastIndex((e) => e.kind === 'message' && e.role === 'user') - // A held message never ran, so the last turn on the page is not its reply. - const turn = settled === 'held' ? [] : entries.slice(lastUser + 1) - return { - ...response, - result: { - ...response.result, - entries: turn, - truncated: settled !== 'held' && truncated && lastUser === -1, - settled - } - } -} - -export function formatSettled(result: ChatTranscript & { settled: ChatSettled }): string { - const status = - result.settled === 'timeout' - ? 'Still working when the wait timed out.' - : result.settled === 'held' - ? 'Your message is held in the queue (the chat was stopped or the send failed). Resume it in Orca.' - : result.settled === 'blocked' - ? 'Blocked: waiting for you to answer an approval or question.' - : 'Done.' - return result.entries.length === 0 ? status : `${formatChatRead(result)}\n${status}` -} - -// The host answers with bare wire codes; say what they mean for this command. -export function explainChatRefusal( - error: unknown, - sessionId: string, - action: 'rename' | 'read' | 'send to' -): unknown { - if (!(error instanceof RuntimeClientError)) { - return error - } - if (error.message === 'agent_session_identity_required') { - return new RuntimeClientError( - 'agent_session_identity_required', - `No chat with session id ${sessionId} on this host.` - ) - } - if (isUnattachedAgentSessionReadRefusal(error)) { - return new RuntimeClientError( - AGENT_SESSION_UNATTACHED_REFUSAL_CODE, - `No open chat with session id ${sessionId} on this host. Find it with \`orca chat list\`.` - ) - } - if ( - error.message === 'structured_agent_session_unsupported' || - error.code === 'method_not_found' - ) { - return new RuntimeClientError( - 'incompatible_runtime', - `This Orca host cannot ${action} chats from the CLI. Update Orca on that host, or check that it has chats turned on.` - ) - } - return error -} diff --git a/src/cli/chat-sessions.ts b/src/cli/chat-sessions.ts deleted file mode 100644 index e88453284f6..00000000000 --- a/src/cli/chat-sessions.ts +++ /dev/null @@ -1,175 +0,0 @@ -import { agentSessionPromptQuestions } from '../shared/agent-session/agent-session-question-answer' -import { isActionableStructuredAgentSessionPrompt } from '../shared/agent-session/view/structured-agent-session-live-turn' -import { isRootAgentJournalItem } from '../shared/agent-session/journal/agent-session-journal-producer' -import type { AgentJournalRenderItem } from '../shared/agent-session/journal/agent-session-journal-types' -import type { - RuntimeWorktreeAgentRow, - RuntimeWorktreePsSummary -} from '../shared/runtime-worktree-contracts' -import { structuredAgentSessionTabId } from '../shared/agent-session/view/structured-agent-session-projection' -import { RuntimeClientError } from './runtime-client' - -const CHAT_HANDLE_PREFIX = structuredAgentSessionTabId('') - -export type ChatSessionEntry = { - session: string - handle: string - worktreeId: string - worktree: string - agentType: string | null - state: RuntimeWorktreeAgentRow['state'] - title: string | null - lastAssistantMessage: string | null - updatedAt: number -} - -export type ChatTranscriptEntry = - | { kind: 'message'; role: string; text: string } - | { kind: 'tool'; name: string; state: string } - | { kind: 'approval' | 'question'; text: string; pending: boolean } - | { kind: 'error'; text: string } - -/** The session id a chat is addressed by, from its id, its handle, or a pane key that names it. */ -export function parseChatSessionSelector(value: string): string { - const withoutHandle = value.startsWith(CHAT_HANDLE_PREFIX) - ? value.slice(CHAT_HANDLE_PREFIX.length) - : value - // A pane key is the chat's handle and its leaf: ":". - const session = withoutHandle.split(':')[0]?.trim() ?? '' - if (session.length === 0) { - throw new RuntimeClientError( - 'invalid_argument', - '--session needs a chat session id or handle; run `orca chat list` to find one.' - ) - } - return session -} - -/** The top-level chats in a worktree summary. Terminal agents and subagents are left out: - * terminals are read with `orca terminal`, and a subagent is reached through its parent chat. */ -export function listChatSessions( - worktrees: readonly RuntimeWorktreePsSummary[] -): ChatSessionEntry[] { - return worktrees.flatMap((worktree) => - worktree.agents - .filter( - (agent) => agent.parentPaneKey === null && agent.paneKey.startsWith(CHAT_HANDLE_PREFIX) - ) - .map((agent) => { - const session = parseChatSessionSelector(agent.paneKey) - return { - session, - handle: structuredAgentSessionTabId(session), - worktreeId: worktree.worktreeId, - worktree: worktree.displayName, - agentType: agent.agentType, - state: agent.state, - title: agent.displayName ?? agent.taskTitle, - lastAssistantMessage: agent.lastAssistantMessage, - updatedAt: agent.updatedAt - } - }) - ) -} - -function messageText(item: AgentJournalRenderItem): string { - if (item.body.kind !== 'message') { - return '' - } - return item.body.blocks - .map((block) => - block.type === 'text' ? block.text : block.type === 'image-ref' ? '[image]' : '' - ) - .filter((text) => text.length > 0) - .join('\n') - .trim() -} - -/** The conversation as a reader follows it: messages, the tools the agent ran, and the - * approvals and questions it raised. Provider status frames and turn markers are left out. */ -export function chatTranscript( - items: readonly AgentJournalRenderItem[], - actionablePromptIds?: readonly string[] -): ChatTranscriptEntry[] { - // An ended agent's unanswered prompts stay pending in the journal but can no longer be answered. - const waiting = (item: AgentJournalRenderItem, state: string): boolean => - state === 'pending' && - isActionableStructuredAgentSessionPrompt(item.itemId, actionablePromptIds) - const entries: ChatTranscriptEntry[] = [] - for (const item of items) { - // Subagents share the journal; their rows belong to their own section, not the chat's turn. - if (!isRootAgentJournalItem(item)) { - continue - } - const body = item.body - if (body.kind === 'message') { - const text = messageText(item) - // Reasoning is the model's scratch work, folded away in the app too. - if (text.length > 0 && body.role !== 'reasoning') { - entries.push({ kind: 'message', role: body.role, text }) - } - } else if (body.kind === 'status' && body.tone === 'error') { - // Start failures and dead agents land as error status rows; without them a wait reads "Done.". - entries.push({ kind: 'error', text: String(body.text) }) - } else if (body.kind === 'tool-call') { - entries.push({ kind: 'tool', name: body.name, state: body.state }) - } else if (body.kind === 'approval') { - entries.push({ - kind: 'approval', - text: body.title, - pending: waiting(item, body.resolution.state) - }) - } else if (body.kind === 'question') { - entries.push({ - kind: 'question', - // Grouped prompts keep their real questions in `questions`; `question` is then a heading. - text: agentSessionPromptQuestions(body) - .map((q) => q.question) - .join('\n'), - pending: waiting(item, body.resolution.state) - }) - } - } - return entries -} - -export function formatChatList(result: { chats: ChatSessionEntry[] }): string { - if (result.chats.length === 0) { - return 'No chats.' - } - return result.chats - .map((chat) => - [ - `${chat.worktree} · ${chat.agentType ?? 'agent'} · ${chat.state}`, - ` session: ${chat.session}`, - ...(chat.title ? [` title: ${chat.title}`] : []), - ...(chat.lastAssistantMessage - ? [` last reply: ${chat.lastAssistantMessage.replace(/\s+/g, ' ').slice(0, 200)}`] - : []) - ].join('\n') - ) - .join('\n\n') -} - -export function formatChatRead(result: { - session: string - entries: ChatTranscriptEntry[] -}): string { - if (result.entries.length === 0) { - return `No messages in ${result.session}.` - } - return result.entries - .map((entry) => { - if (entry.kind === 'message') { - return `${entry.role}: ${entry.text}` - } - if (entry.kind === 'error') { - return `[error] ${entry.text}` - } - if (entry.kind === 'tool') { - return `[tool ${entry.name}: ${entry.state}]` - } - return `[${entry.kind}${entry.pending ? ', waiting for you' : ''}] ${entry.text}` - }) - .join('\n') -} diff --git a/src/cli/command-scoped-flag-help.ts b/src/cli/command-scoped-flag-help.ts index 92e83f0a753..fc5c6c56006 100644 --- a/src/cli/command-scoped-flag-help.ts +++ b/src/cli/command-scoped-flag-help.ts @@ -2,22 +2,7 @@ const FILE_OPEN_FOCUS_HELP = "--focus Bring the user to the file (switches Orca's window to its worktree)" -const CHAT_SESSION_HELP = - '--session Chat session id or handle from chat list or agent launch' - const COMMAND_SCOPED_FLAG_HELP: Record> = { - 'chat rename': { - session: '--session Chat session id from chat list or agent launch' - }, - 'chat list': { - state: '--state Only chats in this state: working, blocked, waiting or done' - }, - 'chat read': { session: CHAT_SESSION_HELP }, - 'chat send': { - session: CHAT_SESSION_HELP, - wait: '--wait Wait until the chat finishes or needs your answer' - }, - 'chat wait': { session: CHAT_SESSION_HELP }, 'extension reload': { 'extension-id': '--extension-id Installed unpacked extension ID; also accepted positionally', diff --git a/src/cli/flag-help-text.ts b/src/cli/flag-help-text.ts index 1e283dd828f..108cf43bf8f 100644 --- a/src/cli/flag-help-text.ts +++ b/src/cli/flag-help-text.ts @@ -36,7 +36,6 @@ export const FLAG_HELP_TEXT: Record = { interrupt: '--interrupt Send as an interrupt-style input when supported', 'keep-open': '--keep-open Keep the worker open after it finishes', 'spec-file': '--spec-file Read the spec from a file or stdin', - 'text-file': '--text-file Read the text from a file or stdin', id: '--id Identifier for a target item or permission', issue: '--issue Linked GitHub issue number', 'gitlab-issue': diff --git a/src/cli/handler-group-manifest.ts b/src/cli/handler-group-manifest.ts index c95a0bec902..cf6e3284a7e 100644 --- a/src/cli/handler-group-manifest.ts +++ b/src/cli/handler-group-manifest.ts @@ -207,7 +207,7 @@ export const HANDLER_GROUPS: readonly HandlerGroup[] = [ }, { name: 'chat', - keys: ['chat rename', 'chat list', 'chat read', 'chat send', 'chat wait'], + keys: ['chat rename'], load: async () => (await import('./handlers/chat.js')).CHAT_HANDLERS }, { diff --git a/src/cli/handlers/chat.test.ts b/src/cli/handlers/chat.test.ts index fb62bc8794f..50005f098b6 100644 --- a/src/cli/handlers/chat.test.ts +++ b/src/cli/handlers/chat.test.ts @@ -1,166 +1,7 @@ import { afterEach, describe, expect, it, vi } from 'vitest' -import { AGENT_SESSION_QUEUED_MESSAGES_RUNTIME_CAPABILITY } from '../../shared/agent-session/agent-session-runtime-capabilities' import { RuntimeClient } from '../runtime-client' import { CHAT_HANDLERS } from './chat' -const SESSION = 'codex_8eb631ab_59c5_41e4_8672_01cfe0a6fd5f' - -function ok(result: unknown) { - return { id: 'req', ok: true, result, _meta: { runtimeId: 'runtime-1' } } -} - -function historyPage( - items: unknown[], - fence: number | undefined = 3, - pageOverrides: Record = {} -) { - return ok({ - ok: true, - page: { - sessionId: SESSION, - epoch: 'e', - fence, - direction: 'tail', - items, - removedItemIds: [], - submissions: [], - window: { - oldest: null, - newest: { epoch: 'e', sequence: 9 }, - nextCursor: { epoch: 'e', sequence: 9 } - }, - ...pageOverrides - } - }) -} - -function workState(fields: Record) { - return ok({ working: false, actionablePrompt: false, ...fields }) -} - -function item(body: unknown) { - return { itemId: 'i', revision: 1, sequence: 1, observedAt: 1, body } -} - -function fakeClient( - responses: unknown[], - capabilities: string[] = [AGENT_SESSION_QUEUED_MESSAGES_RUNTIME_CAPABILITY] -): RuntimeClient & { call: ReturnType } { - const client = new RuntimeClient('/tmp/orca-chat-test', 60_000, null, null) - const call = vi.fn() - for (const response of responses) { - call.mockResolvedValueOnce(response) - } - vi.spyOn(client, 'call').mockImplementation(call) - vi.spyOn(client, 'getCliStatus').mockResolvedValue({ - id: 'status', - ok: true, - result: { - app: { running: true, pid: 123 }, - runtime: { state: 'ready', reachable: true, runtimeId: 'runtime-1', capabilities }, - graph: { state: 'ready' } - }, - _meta: { runtimeId: 'runtime-1' } - }) - return Object.assign(client, { call }) -} - -async function run( - command: string, - flags: Record, - client: RuntimeClient -) { - const logs: string[] = [] - const spy = vi.spyOn(console, 'log').mockImplementation((line: string) => logs.push(line)) - try { - await CHAT_HANDLERS[command]!({ - flags: new Map(Object.entries(flags)), - client, - cwd: '/tmp', - json: true - }) - } finally { - spy.mockRestore() - } - return JSON.parse(logs.join('\n')).result -} - -afterEach(() => { - vi.useRealTimers() - vi.restoreAllMocks() -}) - -describe('chat send', () => { - it('fences a queued message to the chat it read', async () => { - const client = fakeClient([ - workState({ fence: 7 }), - ok({ - ok: true, - replayed: false, - fence: 7, - cursor: { epoch: 'e', sequence: 2 }, - value: { - clientMessageId: 'm-1', - queued: { messageId: 'q-1', position: 1, state: 'waiting' } - } - }) - ]) - const result = await run('chat send', { session: SESSION, text: 'Run the tests' }, client) - const params = client.call.mock.calls[1]![1] - expect(client.call.mock.calls[1]![0]).toBe('agentSession.send') - expect(params).toMatchObject({ - envelope: { sessionId: SESSION, expectedRuntimeFence: 7 }, - delivery: 'queue-if-active', - body: { kind: 'message', role: 'user', blocks: [{ type: 'text', text: 'Run the tests' }] } - }) - expect(result).toEqual({ - session: SESSION, - delivery: 'queued', - clientMessageId: 'm-1', - queuePosition: 1 - }) - }) - - it('with --wait, waits for its queued message to run and returns that turn', async () => { - vi.useFakeTimers() - const client = fakeClient([ - workState({ fence: 7, working: true }), - ok({ - ok: true, - replayed: false, - fence: 7, - cursor: { epoch: 'e', sequence: 2 }, - value: { - clientMessageId: 'm-1', - queued: { messageId: 'q-1', position: 1, state: 'waiting' } - } - }), - // The earlier turn ended, but the queued message has not been handed to the agent yet. - workState({ fence: 7, queuedMessageIds: ['q-1'], nextQueuedMessageId: 'q-1' }), - workState({ fence: 7, working: true, queuedMessageIds: [] }), - workState({ fence: 7, queuedMessageIds: [] }), - historyPage([ - item({ kind: 'message', role: 'user', blocks: [{ type: 'text', text: 'Earlier' }] }), - item({ kind: 'message', role: 'assistant', blocks: [{ type: 'text', text: 'Old reply' }] }), - item({ kind: 'message', role: 'user', blocks: [{ type: 'text', text: 'Run the tests' }] }), - item({ kind: 'message', role: 'assistant', blocks: [{ type: 'text', text: 'All pass' }] }) - ]) - ]) - const pending = run( - 'chat send', - { session: SESSION, text: 'Run the tests', wait: true }, - client - ) - await vi.advanceTimersByTimeAsync(5_000) - const result = await pending - expect(result).toMatchObject({ - delivery: 'queued', - settled: 'done', - entries: [{ kind: 'message', role: 'assistant', text: 'All pass' }] - }) - }) -}) - function chatClient(isRemote: boolean): { client: RuntimeClient; call: ReturnType } { const client = new RuntimeClient('/tmp/orca-chat-test', 60_000, null, null) vi.spyOn(client, 'isRemote', 'get').mockReturnValue(isRemote) diff --git a/src/cli/handlers/chat.ts b/src/cli/handlers/chat.ts index f1c2322f312..f9303026a8d 100644 --- a/src/cli/handlers/chat.ts +++ b/src/cli/handlers/chat.ts @@ -1,64 +1,11 @@ -import { randomUUID } from 'node:crypto' -import type { - AgentSessionMutationResult, - AgentSessionSendResult -} from '../../shared/agent-session/agent-session-wire' -import { AGENT_SESSION_QUEUED_MESSAGES_RUNTIME_CAPABILITY } from '../../shared/agent-session/agent-session-runtime-capabilities' -import type { - RuntimeWorktreePsResult, - RuntimeWorktreeRecord -} from '../../shared/runtime-worktree-contracts' -import { createStructuredAgentSessionOperationId } from '../../shared/agent-session/lifecycle/structured-agent-session-mutation' -import { - structuredAgentSessionMessageSendMutation, - structuredAgentSessionSendBody -} from '../../shared/agent-session/requests/structured-agent-session-send-mutation' -import { AGENT_STATUS_STATES } from '../../shared/agent-status/contracts/agent-status-types' -import { readInjectedAgentSessionId } from '../../shared/agent-session/lifecycle/agent-session-caller-env' -import { - DEFAULT_READ_LIMIT, - chatWorkStateReader, - explainChatRefusal, - formatSettled, - readSettledTurn, - readTranscript, - requireWorkingFlag, - waitForChat -} from '../chat-history' -import { - formatChatList, - formatChatRead, - listChatSessions, - parseChatSessionSelector -} from '../chat-sessions' -import { readTextOrFileFlag } from '../command-body-input' import type { CommandHandler } from '../dispatch' -import { - getOptionalPositiveIntegerFlag, - getOptionalStringFlag, - getPresentStringFlag, - getRequiredStringFlag -} from '../flags' +import { getPresentStringFlag, getRequiredStringFlag } from '../flags' import { printResult } from '../format' -import { getOptionalWorktreeSelector } from '../selectors' -import type { RuntimeClient } from '../runtime-client' import { RuntimeClientError } from '../runtime-client' - -const MAX_READ_LIMIT = 100 +import { readInjectedAgentSessionId } from '../../shared/agent-session/lifecycle/agent-session-caller-env' type ChatRenameResult = { name: string | null; applied: boolean } -// Paired CLIs lack the host's chat read/send capability; refuse before the host does with a bare code. -function readSession(flags: Map, client: RuntimeClient): string { - if (client.isRemote) { - throw new RuntimeClientError( - 'runtime_unsupported', - 'Reading and messaging chats on a paired host is not supported yet. Run this on the host machine.' - ) - } - return parseChatSessionSelector(getRequiredStringFlag(flags, 'session')) -} - export const CHAT_HANDLERS: Record = { 'chat rename': async ({ flags, client, json }) => { const name = getRequiredStringFlag(flags, 'name') @@ -77,120 +24,34 @@ export const CHAT_HANDLERS: Record = { const result = await client .call('agentSession.rename', { sessionId, name }) .catch((error: unknown) => { - throw explainChatRefusal(error, sessionId, 'rename') + throw explainRenameRefusal(error, sessionId) }) if (!result.result.applied) { throw new RuntimeClientError('invalid_argument', 'Not renamed: the name has no visible text.') } printResult(result, json, (value) => `Renamed chat to "${value.name}".`) - }, - 'chat list': async ({ flags, client, cwd, json }) => { - const state = getOptionalStringFlag(flags, 'state') - if (state !== undefined && !AGENT_STATUS_STATES.some((known) => known === state)) { - throw new RuntimeClientError( - 'invalid_argument', - `--state must be one of: ${AGENT_STATUS_STATES.join(', ')}.` - ) - } - const selector = await getOptionalWorktreeSelector(flags, 'worktree', cwd, client) - const shown = selector - ? ( - await client.call<{ worktree: RuntimeWorktreeRecord }>('worktree.show', { - worktree: selector - }) - ).result.worktree - : undefined - const result = await client.call('worktree.ps', {}) - // A local and an SSH copy can share an id; the host tells them apart when both name one. - const worktrees = result.result.worktrees.filter( - (w) => - !shown || - (w.worktreeId === shown.id && - (w.hostId === undefined || shown.hostId === undefined || w.hostId === shown.hostId)) - ) - const chats = listChatSessions(worktrees).filter((chat) => !state || chat.state === state) - printResult({ ...result, result: { chats } }, json, formatChatList) - }, - 'chat read': async ({ flags, client, json }) => { - const session = readSession(flags, client) - const limit = getOptionalPositiveIntegerFlag(flags, 'limit') ?? DEFAULT_READ_LIMIT - if (limit > MAX_READ_LIMIT) { - throw new RuntimeClientError('invalid_argument', `--limit must be at most ${MAX_READ_LIMIT}.`) - } - printResult(await readTranscript(client, session, limit), json, formatChatRead) - }, - 'chat wait': async ({ flags, client, json }) => { - const session = readSession(flags, client) - const timeoutMs = getOptionalPositiveIntegerFlag(flags, 'timeout-ms') - const settled = await waitForChat(chatWorkStateReader(client, session), timeoutMs) - printResult(await readSettledTurn(client, session, settled), json, formatSettled) - }, - 'chat send': async ({ flags, client, cwd, json }) => { - const session = readSession(flags, client) - const text = (await readTextOrFileFlag(flags, cwd, 'text', true)) ?? '' - if (text.trim().length === 0) { - throw new RuntimeClientError('invalid_argument', 'The message must not be empty.') - } - const wait = flags.has('wait') - const timeoutMs = getOptionalPositiveIntegerFlag(flags, 'timeout-ms') - const status = await client.getCliStatus() - if (!status.result.runtime.reachable) { - throw new RuntimeClientError( - 'runtime_unavailable', - 'Orca could not reach the execution host, so nothing was sent. Retry once it is reachable.' - ) - } - // A working chat holds a queued message until its turn ends; older hosts get an ordinary send. - const queue = - status.result.runtime.capabilities?.includes( - AGENT_SESSION_QUEUED_MESSAGES_RUNTIME_CAPABILITY - ) === true - // The send is fenced to the session's current runtime, so a restarted chat refuses it. - const readWorkState = chatWorkStateReader(client, session) - const state = await readWorkState() - if (wait) { - requireWorkingFlag(state) - } - const { fence } = state - if (fence === undefined) { - throw new RuntimeClientError( - 'agent_session_not_running', - `Chat ${session} has no running session on this host, so nothing was sent. Open it in Orca and retry.` - ) - } - const mutation = structuredAgentSessionMessageSendMutation({ - sessionId: session, - clientOperationId: createStructuredAgentSessionOperationId(randomUUID), - expectedRuntimeFence: fence, - body: structuredAgentSessionSendBody(text, []), - ...(queue ? { delivery: 'queue-if-active' as const } : {}) - }) - const result = await client.call>( - 'agentSession.send', - mutation - ) - if (!result.result.ok) { - const { code, message } = result.result.refusal - throw explainChatRefusal(new RuntimeClientError(code, message), session, 'send to') - } - const value = result.result.value - const sent = { - session, - delivery: 'queued' in value ? ('queued' as const) : ('submitted' as const), - clientMessageId: value.clientMessageId, - ...('queued' in value ? { queuePosition: value.queued.position } : {}) - } - if (!wait) { - printResult({ ...result, result: sent }, json, (r) => - r.delivery === 'queued' - ? `Queued for ${r.session}; it is sent when the current turn ends.` - : `Sent to ${r.session}.` - ) - return - } - const queuedId = 'queued' in value ? value.queued.messageId : undefined - const settled = await waitForChat(readWorkState, timeoutMs, queuedId) - const turn = await readSettledTurn(client, session, settled) - printResult({ ...turn, result: { ...sent, ...turn.result } }, json, formatSettled) } } + +// The host answers with bare wire codes; say what they mean for this command. +function explainRenameRefusal(error: unknown, sessionId: string): unknown { + if (!(error instanceof RuntimeClientError)) { + return error + } + if (error.message === 'agent_session_identity_required') { + return new RuntimeClientError( + 'agent_session_identity_required', + `No chat with session id ${sessionId} on this host.` + ) + } + if ( + error.message === 'structured_agent_session_unsupported' || + error.code === 'method_not_found' + ) { + return new RuntimeClientError( + 'incompatible_runtime', + 'This Orca host cannot rename chats from the CLI. Update Orca on that host, or check that it has chats turned on.' + ) + } + return error +} diff --git a/src/cli/root-help-text-primary.ts b/src/cli/root-help-text-primary.ts index 502443d3d69..0b0415336b9 100644 --- a/src/cli/root-help-text-primary.ts +++ b/src/cli/root-help-text-primary.ts @@ -121,10 +121,6 @@ export const ROOT_HELP_TEXT_PRIMARY = [ ' agent launch Start an agent in a worktree, as a chat or terminal per your settings', '', 'Chats:', - ' chat list List agent chats with their workspace, state and last reply', - ' chat read Read the latest messages of an agent chat', - ' chat send Send a message to an agent chat', - ' chat wait Wait until an agent chat finishes or needs your answer', ' chat rename Rename a chat tab', '', 'Files:', diff --git a/src/cli/specs/chat.ts b/src/cli/specs/chat.ts index ae73e34d6ed..28fb41556dd 100644 --- a/src/cli/specs/chat.ts +++ b/src/cli/specs/chat.ts @@ -15,57 +15,5 @@ export const CHAT_COMMAND_SPECS: CommandSpec[] = [ 'orca chat rename --name "Migrate auth to OAuth"', 'orca chat rename --session --name "Release notes" --json' ] - }, - { - path: ['chat', 'list'], - summary: 'List agent chats with their workspace, state and last reply', - usage: 'orca chat list [--worktree ] [--state ] [--json]', - allowedFlags: [...GLOBAL_FLAGS, 'worktree', 'state'], - notes: [ - 'Lists the top-level chats Orca shows as agent tabs. Terminal agents are read with `orca terminal`.', - '--state keeps chats in one state: working, blocked, waiting or done.' - ], - examples: ['orca chat list --json', 'orca chat list --state working'] - }, - { - path: ['chat', 'read'], - summary: 'Read the latest messages of an agent chat', - usage: 'orca chat read --session [--limit ] [--json]', - allowedFlags: [...GLOBAL_FLAGS, 'session', 'limit'], - notes: [ - '--session takes the session id or handle from `orca chat list` or `orca agent launch`.', - '--limit is the number of entries shown (messages, tool calls, approvals and questions); the default is 20 and the maximum 100.', - 'Approvals and questions waiting for an answer are marked "waiting for you".' - ], - examples: ['orca chat read --session codex_1234 --limit 10'] - }, - { - path: ['chat', 'send'], - summary: 'Send a message to an agent chat', - usage: - 'orca chat send --session (--text | --text-file ) [--wait] [--timeout-ms ] [--json]', - allowedFlags: [...GLOBAL_FLAGS, 'session', 'text', 'text-file', 'wait', 'timeout-ms'], - notes: [ - 'Sends the text as your message, as if typed in the chat. --text-file - reads it from stdin.', - 'While the agent is working, hosts with a message queue hold the message until the turn ends; the result says "queued".', - '--wait returns once the chat finishes or waits for your answer, with what it did since your message. It exits 1 when the wait times out or the message is held in a stopped queue.', - 'A chat that restarted since it was read refuses the send; read it again and resend.' - ], - examples: [ - 'orca chat send --session codex_1234 --text "Run the tests, then summarize failures"', - 'orca chat send --session codex_1234 --text "Run the tests" --wait --timeout-ms 600000 --json', - 'cat notes.md | orca chat send --session codex_1234 --text-file -' - ] - }, - { - path: ['chat', 'wait'], - summary: 'Wait until an agent chat finishes or needs your answer', - usage: 'orca chat wait --session [--timeout-ms ] [--json]', - allowedFlags: [...GLOBAL_FLAGS, 'session', 'timeout-ms'], - notes: [ - 'Settles on the chat\'s own turn, in `chat list` words: "done", or "blocked" on an approval or question. Background agents it started can still show as working or waiting in `chat list`. "held" and "timeout" exit 1.', - 'Prints what the chat did since the last user message. A wait that times out exits 1.' - ], - examples: ['orca chat wait --session codex_1234 --timeout-ms 600000 --json'] } ] diff --git a/src/main/native-chat/agent-session-journal/input-order/journal-newest-turn.ts b/src/main/native-chat/agent-session-journal/input-order/journal-newest-turn.ts index a7b3d7d8781..5b6e6137a80 100644 --- a/src/main/native-chat/agent-session-journal/input-order/journal-newest-turn.ts +++ b/src/main/native-chat/agent-session-journal/input-order/journal-newest-turn.ts @@ -50,13 +50,6 @@ export class JournalNewestTurn { return this.root?.latest.item ?? null } - /** The newest turn or clear, when it is a turn still running. */ - running(): { item: AgentJournalRenderItem; turnId: string } | null { - const item = this.read() - const turn = item ? readAgentJournalTurn(item.body) : null - return item && turn?.state === 'running' ? { item, turnId: turn.turnId } : null - } - /** Releasing Resume marks must be written after this position. */ failure(): number { const winner = this.root?.winner diff --git a/src/main/native-chat/agent-session-journal/journal-store.ts b/src/main/native-chat/agent-session-journal/journal-store.ts index 7ad19f089e6..cb22b39021f 100644 --- a/src/main/native-chat/agent-session-journal/journal-store.ts +++ b/src/main/native-chat/agent-session-journal/journal-store.ts @@ -15,6 +15,7 @@ import type { import { agentJournalItemKey } from '../../../shared/agent-session/journal/agent-session-journal-item-key' import type { AgentSessionContextUsage } from '../../../shared/agent-session/agent-session-context-usage' import { latestStructuredAgentContextFacts } from '../../../shared/agent-session/structured-agent-session-context-usage' +import { runningStructuredAgentSessionTurnItemBySequence } from '../../../shared/agent-session/view/structured-agent-session-live-turn' import { readAgentJournalTurn } from '../../../shared/agent-session/lifecycle/agent-session-turn-record' import { journalWroteBeforeOpen } from './replay/journal-cursor' import type { JournalHostDatabase } from './database/journal-host-database' @@ -229,8 +230,9 @@ export class AgentSessionJournal { hasRecordedItem = (itemId: string): boolean => this.state.items.has(itemId) || this.state.tombstones.has(itemId) - /** Read from the per-row turn index so hot readers never rescan history; host projection decides current work. */ - runningTurn = () => journalInputOrder(this.state).newestTurn.running() + /** Read the journal's running turn without allocating a snapshot; host projection decides current work. */ + runningTurn = (): { item: journal.AgentJournalRenderItem; turnId: string } | null => + runningStructuredAgentSessionTurnItemBySequence(this.state.items.values()) newestTurn = (): journal.AgentJournalTurnLifecycle | null => readAgentJournalTurn(journalInputOrder(this.state).newestTurn.read()?.body) diff --git a/src/main/native-chat/agent-session-wire/host/structured-agent-session-host.ts b/src/main/native-chat/agent-session-wire/host/structured-agent-session-host.ts index 9b5eb8e25a1..1d4ed8ab39b 100644 --- a/src/main/native-chat/agent-session-wire/host/structured-agent-session-host.ts +++ b/src/main/native-chat/agent-session-wire/host/structured-agent-session-host.ts @@ -62,8 +62,7 @@ export class StructuredAgentSessionHost { private readonly lifetime: HostTypes.StructuredAgentSessionConversationLifetime private readonly conversationDelivery: conversation.StructuredAgentSessionConversationDelivery private readonly eventRecovery: StructuredAgentSessionEventRecovery - /** History and work-state reads, which open a chat at rest without starting its agent. */ - readonly backgroundTasks: BackgroundTasks + private readonly backgroundTasks: BackgroundTasks readonly restartResume: StructuredAgentSessionRestartResume readonly usageLimitContinue: ReturnType diff --git a/src/main/native-chat/agent-session-wire/publication/subscribers/structured-agent-session-background-task-channel.ts b/src/main/native-chat/agent-session-wire/publication/subscribers/structured-agent-session-background-task-channel.ts index f5ec1229325..65d513d7ff1 100644 --- a/src/main/native-chat/agent-session-wire/publication/subscribers/structured-agent-session-background-task-channel.ts +++ b/src/main/native-chat/agent-session-wire/publication/subscribers/structured-agent-session-background-task-channel.ts @@ -6,9 +6,6 @@ import type { AgentSessionHistoryRequest, AgentSessionHistoryResult } from '../../../../../shared/agent-session/agent-session-wire' -import type { AgentSessionWorkState } from '../../../../../shared/agent-session/agent-session-work-state' -import { AgentSessionRefusalError } from '../../../../../shared/agent-session/failures/agent-session-wire-refusals' -import { AGENT_SESSION_NOT_ATTACHED } from '../../operations/structured-agent-session-mutation-refusals' import type { AgentSessionBackgroundTaskState } from '../../../../../shared/agent-session/agent-session-background-task-wire' import type { AgentChildWorkView } from '../../../../../shared/agent-status/child-work/agent-status-child-work-view' import { structuredChildWorkLegacyTasks } from '../../../../../shared/agent-session/structured-agent-session-child-work-legacy' @@ -88,31 +85,6 @@ export class StructuredAgentSessionBackgroundTaskChannel { } } - /** What a waiter decides from, from the sources history pages state, without building a page. */ - async workState(sessionId: string): Promise { - await this.conversation(sessionId) - const work = hostStructuredAgentSessionCurrentWork( - { store: this.deps.store, sessions: this.sessions }, - sessionId - ) - if (!work) { - throw new AgentSessionRefusalError(AGENT_SESSION_NOT_ATTACHED) - } - const queue = this.subscribers.readQueuePublication(sessionId) - const fence = this.deps.store.getConversation(sessionId)?.currentFence - return { - working: work.working(), - actionablePrompt: work.hasActionablePrompt(), - ...(queue !== undefined - ? { - queuedMessageIds: (queue.queuedMessages ?? []).map((card) => card.messageId), - nextQueuedMessageId: queue.nextQueuedMessageId - } - : {}), - ...(fence !== undefined ? { fence } : {}) - } - } - async currentRequests(request: AgentSessionCurrentRequestsRead) { await this.conversation(request.sessionId) const work = hostStructuredAgentSessionCurrentWork( diff --git a/src/main/runtime/agent-session/structured-agent-session-chat-wait-work-state.test.ts b/src/main/runtime/agent-session/structured-agent-session-chat-wait-work-state.test.ts deleted file mode 100644 index 31d488df6e2..00000000000 --- a/src/main/runtime/agent-session/structured-agent-session-chat-wait-work-state.test.ts +++ /dev/null @@ -1,124 +0,0 @@ -import { afterEach, describe, expect, it, vi } from 'vitest' -import type { RpcResponse } from '../rpc/core' -import { AGENT_SESSION_QUEUED_MESSAGES_RUNTIME_CAPABILITY } from '../../../shared/agent-session/agent-session-runtime-capabilities' -import { RuntimeClient, RuntimeClientError } from '../../../cli/runtime-client' -import { CHAT_HANDLERS } from '../../../cli/handlers/chat' -import { - openStructuredCodexRpcHarness, - SESSION, - THREAD, - TURN, - type StructuredCodexRpcHarness -} from '../rpc/methods/structured-sessions/structured-codex-session-rpc-test-harness' - -const NEWEST_MARKER = 'newest-row-marker' - -let harness: StructuredCodexRpcHarness | undefined - -afterEach(async () => { - vi.useRealTimers() - vi.restoreAllMocks() - await harness?.dispose() - harness = undefined -}) - -/** The CLI's own client, answered by the host's dispatcher in the local CLI's caller context. */ -function cliClient(open: StructuredCodexRpcHarness) { - const client = new RuntimeClient('/tmp/orca-chat-wait-test', 60_000, null, null) - vi.spyOn(client, 'getCliStatus').mockResolvedValue({ - id: 'status', - ok: true, - result: { - app: { running: true, pid: 1 }, - runtime: { - state: 'ready', - reachable: true, - runtimeId: 'runtime-1', - capabilities: [AGENT_SESSION_QUEUED_MESSAGES_RUNTIME_CAPABILITY] - }, - graph: { state: 'ready' } - }, - _meta: { runtimeId: 'runtime-1' } - }) - vi.spyOn(client, 'call').mockImplementation(async (method: string, params?: unknown) => { - const response: RpcResponse = await open.call(method, params) - if (!response.ok) { - throw new RuntimeClientError(response.error.code, response.error.message) - } - return response - }) - return client -} - -function agentMessage(id: string, text: string): void { - harness!.codex.notify('item/completed', { - threadId: THREAD, - turnId: TURN, - item: { type: 'agentMessage', id, text } - }) -} - -/** A running turn holding `rows` earlier replies and one newest reply of `newestBytes`. */ -async function seedRunningTurn(rows: number, newestBytes: number): Promise { - harness!.codex.notify('turn/started', { threadId: THREAD, turn: { id: TURN } }) - for (let index = 0; index < rows; index += 1) { - agentMessage(`older-${index}`, `Older reply ${index}.`) - } - agentMessage('newest', `${NEWEST_MARKER} ${'x'.repeat(newestBytes)}`) - await vi.waitFor(async () => { - const page = await harness!.call('agentSession.history', { - sessionId: SESSION, - direction: 'tail', - limit: 1 - }) - expect(JSON.stringify(page)).toContain(NEWEST_MARKER) - }) -} - -/** Lets the command's next checks run until it prints, then surfaces its result or failure. */ -async function settle(command: Promise, logs: string[]): Promise { - for (let second = 0; second < 10 && logs.length === 0; second += 1) { - await vi.advanceTimersByTimeAsync(1_000) - } - await command -} - -describe('chat send --wait over the real host', () => { - it('returns the completed turn when the host lacks agentSession.workState', async () => { - harness = await openStructuredCodexRpcHarness( - [], - {}, - { localCli: true, omitMethods: ['agentSession.workState'] } - ) - await harness.ok('agentSession.create', harness.createIntentParams()) - const client = cliClient(harness) - const logs: string[] = [] - vi.spyOn(console, 'log').mockImplementation((line: string) => logs.push(line)) - vi.useFakeTimers({ toFake: ['setTimeout', 'clearTimeout'] }) - const sending = CHAT_HANDLERS['chat send']!({ - flags: new Map([ - ['session', SESSION], - ['text', 'Run the tests'], - ['wait', true] - ]), - client, - cwd: '/tmp', - json: true - }) - await vi.waitFor(() => - expect(harness!.codex.live().calls.map((c) => c.method)).toContain('turn/start') - ) - await seedRunningTurn(1, 16) - await vi.advanceTimersByTimeAsync(1_000) - harness.codex.notify('turn/completed', { - threadId: THREAD, - turn: { id: TURN, status: 'completed' } - }) - await settle(sending, logs) - expect(JSON.parse(logs.join('\n')).result).toMatchObject({ - delivery: 'submitted', - settled: 'done', - entries: [{ text: 'Older reply 0.' }, { text: `${NEWEST_MARKER} ${'x'.repeat(16)}` }] - }) - }) -}) diff --git a/src/main/runtime/rpc/methods/structured-sessions/structured-agent-session.ts b/src/main/runtime/rpc/methods/structured-sessions/structured-agent-session.ts index 9d7d6ea608d..6ae64ac2c10 100644 --- a/src/main/runtime/rpc/methods/structured-sessions/structured-agent-session.ts +++ b/src/main/runtime/rpc/methods/structured-sessions/structured-agent-session.ts @@ -45,7 +45,6 @@ import { CreateParams, CreateSupportParams, HistoryParams, - WorkStateParams, RewindParams, SendParams, SleepWorkspaceParams, @@ -192,16 +191,6 @@ export const STRUCTURED_AGENT_SESSION_METHODS = [ ) } }), - // Waiters poll this instead of history: no history page or running-turn scan, and no transcript in the reply. - defineMethod({ - name: 'agentSession.workState', - permission: 'workspace', - params: WorkStateParams, - handler: async (params, ctx) => - (await requireInstalledHost(ctx, params.sessionId)).backgroundTasks.workState( - params.sessionId - ) - }), STRUCTURED_AGENT_SESSION_SUBSCRIBE_METHOD, defineMethod({ name: 'agentSession.unsubscribe', diff --git a/src/main/runtime/rpc/methods/structured-sessions/structured-codex-session-rpc-test-harness.ts b/src/main/runtime/rpc/methods/structured-sessions/structured-codex-session-rpc-test-harness.ts index 71071e1d5bb..114754dc4f4 100644 --- a/src/main/runtime/rpc/methods/structured-sessions/structured-codex-session-rpc-test-harness.ts +++ b/src/main/runtime/rpc/methods/structured-sessions/structured-codex-session-rpc-test-harness.ts @@ -181,13 +181,11 @@ export type StructuredCodexRpcHarness = { export async function openStructuredCodexRpcHarness( clientCapabilities: readonly RuntimeCapability[] = DEFAULT_CLIENT_CAPABILITIES, /** What the transport stamps, as a paired device's authenticated socket does. */ - transport: { clientId?: string; pairedDeviceId?: string } = {}, - /** The local CLI names no client kind; an older host lacks newer methods. */ - options: { localCli?: boolean; omitMethods?: readonly string[] } = {} + transport: { clientId?: string; pairedDeviceId?: string } = {} ): Promise { const client = { clientId: 'device-a', - ...(options.localCli ? {} : { clientKind: 'runtime' as const }), + clientKind: 'runtime' as const, clientCapabilities, ...transport } @@ -249,9 +247,7 @@ export async function openStructuredCodexRpcHarness( } const dispatcher = new RpcDispatcher({ runtime: runtime as unknown as OrcaRuntimeService, - methods: STRUCTURED_AGENT_SESSION_METHODS.filter( - (method) => !options.omitMethods?.includes(method.name) - ) + methods: STRUCTURED_AGENT_SESSION_METHODS }) const call: StructuredCodexRpcHarness['call'] = async (method, params) => { const replies: RpcResponse[] = [] diff --git a/src/shared/agent-session/agent-session-work-state.ts b/src/shared/agent-session/agent-session-work-state.ts deleted file mode 100644 index 8ac90393f9d..00000000000 --- a/src/shared/agent-session/agent-session-work-state.ts +++ /dev/null @@ -1,12 +0,0 @@ -/** `agentSession.workState`: the facts a waiter decides from, read without a history page. */ -export type AgentSessionWorkState = { - working: boolean - /** A question or approval is waiting for the user. */ - actionablePrompt: boolean - /** Absent = the host makes no queue claim. */ - queuedMessageIds?: string[] - /** With queuedMessageIds, the next unblocked send; null while held. */ - nextQueuedMessageId?: string | null - /** The session's current runtime fence, as history pages state it. */ - fence?: number -} diff --git a/src/shared/rpc-contract/rpc-params-catalog.generated.ts b/src/shared/rpc-contract/rpc-params-catalog.generated.ts index 66f1f50c1b1..7d58d1a15df 100644 --- a/src/shared/rpc-contract/rpc-params-catalog.generated.ts +++ b/src/shared/rpc-contract/rpc-params-catalog.generated.ts @@ -670,8 +670,7 @@ import { SleepWorkspaceParams, SubscribeTurnCompletionsParams, ThreadGoalParams, - UnsubscribeParams, - WorkStateParams + UnsubscribeParams } from './structured-agent-session-params' import { RespondToQuestionParams } from './structured-agent-session-question-params' import { @@ -880,7 +879,6 @@ export const RPC_PARAMS_BY_METHOD = { 'agentSession.subscribeTurnCompletions': SubscribeTurnCompletionsParams, 'agentSession.threadGoal': ThreadGoalParams, 'agentSession.unsubscribe': UnsubscribeParams, - 'agentSession.workState': WorkStateParams, 'agentSessionAttachment.read': AttachmentReadParams, 'agentSessionAttachment.readTextChunk': AttachmentTextReadParams, 'agentSessionAttachment.uploadAbort': AttachmentUploadIdParams, diff --git a/src/shared/rpc-contract/structured-agent-session-params.ts b/src/shared/rpc-contract/structured-agent-session-params.ts index fc31b5e97c1..208bc6acba3 100644 --- a/src/shared/rpc-contract/structured-agent-session-params.ts +++ b/src/shared/rpc-contract/structured-agent-session-params.ts @@ -335,8 +335,6 @@ export const HistoryParams = z }) .strict() -export const WorkStateParams = z.object({ sessionId: SessionId }).strict() - // Not strict: an older host ignores these params entirely, and this host must ignore a newer client's. export const SubscribeTurnCompletionsParams = z.object({ includePrompts: z.boolean().optional() })