Merge origin/main into ota-c7-5b-document-factory (OTA phase C, C7.5b)

Main at 2d697a4012, which carries #21908's re-pin of the session route's module
count. One conflict, the pin itself, plus `27a0889dcf`'s relay test arriving
unconflicted.

Resolved by keeping this branch's `SESSION_ROUTE_MODULES = 4284` and folding
#21908's account into the one explanation the docstring already holds. Re-measured
on the merged tree with the four postinstall generators run first rather than
carried over: 4284 modules, 934 local, with
`src/mobile-web-shell/bridge/bridge-haptics-notify.ts` read out of the closure by
name rather than inferred from the total.

What #21908 adds to the account, and what this branch could not say on its own:
C7.10 item E (#21864) and mermaid (#21871) were each green against a main that
lacked the other, which is how main came to hold 4323 while measuring 4324. The
module reaches this branch's 4284 by that same route and not by anything this
branch did. The constant's own comment is now the short form and points at the
docstring, so the two explanations read once rather than twice.

Nothing under `mobile/` moves in this merge. It changes exactly two files here,
the closure test and `src/relay/omp-fresh-launch-environment.test.ts`, so the
mobile suite is not re-run; mobile tsc is, and is clean.

Claude-Session: https://claude.ai/code/session_01JNnE9qzUZMMnqpZWCqM3nb

config/scripts/mobile-web-app-session-terminal-closure.test.mjs

@@ -24,11 +24,11 @@ import {
  *   modules        4324 -> 4284   (-40)
  *   local modules   974 ->  934   (-40)
  *
- * The extra module is `src/mobile-web-shell/bridge/bridge-haptics-notify.ts`, which C7.10 item E
- * put on the session route after the +3 below was recorded, so main itself reads 4324 against the
- * 4323 it holds and #21908 re-pins it there. It is in this branch's 4284 for the same reason and
- * by the same route, not as anything this branch did: `haptics.web.ts` was already in the closure
- * and the bridge module joins it.
+ * The extra module is `src/mobile-web-shell/bridge/bridge-haptics-notify.ts`, which
+ * `haptics.web.ts` reaches. C7.10 item E (#21864) and mermaid (#21871) were each green against a
+ * main that lacked the other, so main held 4323 while measuring 4324, and #21908 re-pinned it
+ * there. It is in this branch's 4284 by that same route rather than by anything this branch did:
+ * `haptics.web.ts` was already in the closure and the bridge module joins it.
  *
  * Which is the point of re-measuring rather than summing. A merged number arrived at as
  * -40 plus +3 would have read 4283 and been wrong about a module neither side of the merge moved.
@@ -136,10 +136,9 @@ const MERMAID_PAGE_ENGINE = 'src/components/pr-sidebar/mermaid-page-engine.gener
 const MERMAID_PACKAGE = 'node_modules/mermaid/'
 
 /**
- * The module list on the merge, recorded at the base in the docstring above.
- *
- * It carries both effects at once -- the factory replacing the document's source modules and
- * mermaid's three -- which is why it is neither side's own number.
+ * The module list on the merge, recorded at the base in the docstring above, which is where the
+ * three things inside it are accounted for: the factory replacing the document's source modules,
+ * mermaid's three, and the haptics notify module #21908 pins on main.
  */
 const SESSION_ROUTE_MODULES = 4284
 

src/relay/omp-fresh-launch-environment.test.ts

@@ -91,7 +91,7 @@ it('prepares the execution host OMP config and status extension for a guarded la
     source.mockReturnValue(false)
     expect(
       await augment.mock.calls[1][0]({ id: 'other', shell: '/bin/bash', env: {}, command: 'codex' })
-    ).toEqual({})
+    ).toEqual({ ORCA_OPENCODE_AGENT: 'opencode' })
   } finally {
     runtime.stop()
     dispatcher.dispose()