diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
index 1c3812eae..856ebf86f 100644
--- a/.github/workflows/release.yml
+++ b/.github/workflows/release.yml
@@ -626,7 +626,7 @@ jobs:
           make_latest: ${{ needs.release-info.outputs.prerelease == 'true' && 'false' || 'true' }}
           body: |
             > [!WARNING]
-            > **Upgrading across major versions?** Database schema and configuration are only compatible within a major version, and the one-click self-update refuses cross-major targets (releases before 3.0 lack this gate — do not use their built-in update to reach a newer major). Deploy the new major fresh (new `.runtime/` data directory) and re-import or re-authorize provider accounts and client keys. See the [upgrade notes](https://github.com/${{ github.repository }}/blob/main/deploy/README.md#镜像升级与源码构建).
+            > **Upgrading across major versions?** Online upgrades across major versions are not supported. Deploy the new major fresh with a new `.runtime/` data directory, then re-import or re-authorize provider accounts and client keys. See the [upgrade notes](https://github.com/${{ github.repository }}/blob/main/deploy/README.md#镜像升级与源码构建).
 
             ## 📥 Installation
 
diff --git a/README.md b/README.md
index 57ca2e40d..e581911eb 100644
--- a/README.md
+++ b/README.md
@@ -168,10 +168,8 @@ docker compose -f deploy/compose.yaml up -d
 > 运行状态。
 
 > [!WARNING]
-> 跨大版本不是原地升级：数据库 schema 与配置只在同一大版本内兼容，内置一键自更新
-> 会拒绝跨大版本目标（3.0 之前的旧版没有这道门禁，不要用旧版界面的一键更新跨大版本）。
-> 升级到新的大版本需要全新部署（全新 `.runtime/` 数据目录），并重新导入或重新授权
-> Provider 账号与客户端 Key。
+> 不支持跨大版本在线升级。升级到新的大版本时，请使用全新的 `.runtime/` 数据目录重新部署，
+> 并重新导入或重新授权 Provider 账号与客户端 Key。
 
 Compose 默认只绑定 `127.0.0.1`。公网接入应使用 HTTPS 反向代理，转发 WebSocket
 upgrade 与真实客户端 IP；不要暴露 PostgreSQL 或 Redis。
diff --git a/backend/crates/gateway-store/tests/postgres/mod.rs b/backend/crates/gateway-store/tests/postgres/mod.rs
index f37bf92dc..2845a4c8f 100644
--- a/backend/crates/gateway-store/tests/postgres/mod.rs
+++ b/backend/crates/gateway-store/tests/postgres/mod.rs
@@ -58,30 +58,6 @@ impl TestDatabase {
             .execute(&mut *migration)
             .await
             .expect("apply terminal migration");
-        sqlx::raw_sql(include_str!(
-            "../../../../migrations/0002_drop_opaque_response_id_index.sql"
-        ))
-        .execute(&mut *migration)
-        .await
-        .expect("apply opaque response ID migration");
-        sqlx::raw_sql(include_str!(
-            "../../../../migrations/0003_store_opaque_response_ids_as_bytes.sql"
-        ))
-        .execute(&mut *migration)
-        .await
-        .expect("apply opaque response storage migration");
-        sqlx::raw_sql(include_str!(
-            "../../../../migrations/0004_add_model_request_service_tier.sql"
-        ))
-        .execute(&mut *migration)
-        .await
-        .expect("apply model request service tier migration");
-        sqlx::raw_sql(include_str!(
-            "../../../../migrations/0005_allow_quota_exhausted_cooldown.sql"
-        ))
-        .execute(&mut *migration)
-        .await
-        .expect("apply quota exhausted cooldown migration");
         migration.commit().await.expect("commit terminal migration");
         Some(Self {
             admin,
@@ -177,25 +153,19 @@ async fn connect_and_migrate_should_apply_all_migrations_once_and_reopen_cleanly
     .expect("drop migration test database");
     admin.close().await;
 
-    assert_eq!((first_table_count, migration_count), (8, 5));
+    assert_eq!((first_table_count, migration_count), (8, 1));
     assert_eq!(response_id_types, ["bytea", "bytea"]);
     assert!(!raw_response_id_index_exists);
 }
 
 #[test]
 fn migrations_should_leave_transaction_ownership_to_sqlx() {
-    let transaction_statements = [
-        include_str!("../../../../migrations/0001_initial.sql"),
-        include_str!("../../../../migrations/0002_drop_opaque_response_id_index.sql"),
-        include_str!("../../../../migrations/0003_store_opaque_response_ids_as_bytes.sql"),
-        include_str!("../../../../migrations/0004_add_model_request_service_tier.sql"),
-        include_str!("../../../../migrations/0005_allow_quota_exhausted_cooldown.sql"),
-    ]
-    .into_iter()
-    .flat_map(str::lines)
-    .map(str::trim)
-    .filter(|line| matches!(*line, "begin;" | "commit;"))
-    .count();
+    let transaction_statements = [include_str!("../../../../migrations/0001_initial.sql")]
+        .into_iter()
+        .flat_map(str::lines)
+        .map(str::trim)
+        .filter(|line| matches!(*line, "begin;" | "commit;"))
+        .count();
 
     assert_eq!(transaction_statements, 0);
 }
diff --git a/backend/migrations/.frozen-sha256 b/backend/migrations/.frozen-sha256
index 5bb9bcacb..616b5bf49 100644
--- a/backend/migrations/.frozen-sha256
+++ b/backend/migrations/.frozen-sha256
@@ -1,5 +1 @@
-0554fed953bd1440b50bc4033dc8bad96b51e132c57c2ccf1553728a991d5cb2  0001_initial.sql
-0ae3945c02461347ab8dcfc64821027cd2366505778e0450d71bd82772c4e99d  0002_drop_opaque_response_id_index.sql
-e608ed89991f9d6317491714447a00a61ede90caca4fb030eb30695654afb664  0003_store_opaque_response_ids_as_bytes.sql
-df4e02cf2335733326c12429ced4cbd0b35b8a22b3fc24981c3c0549477d7c43  0004_add_model_request_service_tier.sql
-fe8f36913c6db7053148db450f05e008dab6cdb9ca8cdcfcb22d97ad9a87bf49  0005_allow_quota_exhausted_cooldown.sql
+a4681a5c040e80c5234e879293fc34972080e5251dfbc7c40e1374f8e4170a4c  0001_initial.sql
diff --git a/backend/migrations/0001_initial.sql b/backend/migrations/0001_initial.sql
index 0b1ac07c6..13f4fcbc3 100644
--- a/backend/migrations/0001_initial.sql
+++ b/backend/migrations/0001_initial.sql
@@ -166,7 +166,14 @@ create table provider_accounts (
     )
   ),
   constraint provider_accounts_cooldown_ck check (
-    (availability = 'cooldown') = (cooldown_until is not null)
+    (
+      availability != 'cooldown'
+      or cooldown_until is not null
+    )
+    and (
+      cooldown_until is null
+      or availability in ('cooldown', 'quota_exhausted')
+    )
   ),
   constraint provider_accounts_time_ck check (
     created_at <= updated_at
@@ -208,6 +215,7 @@ create table model_requests (
   requested_model_id text not null,
   provider_kind text,
   upstream_model_id text,
+  service_tier text,
   provider_account_id text,
   provider_account_ref text,
   upstream_transport text,
@@ -220,9 +228,11 @@ create table model_requests (
   outcome text not null default 'running',
   client_status_code integer,
   upstream_status_code integer,
-  client_response_id text,
+  -- Provider response IDs are opaque JSON strings. Bytes preserve embedded NUL
+  -- and avoid imposing PostgreSQL text-index limits on protocol handles.
+  client_response_id bytea,
   upstream_request_id text,
-  upstream_response_id text,
+  upstream_response_id bytea,
   error_kind text,
   provider_error_code text,
   error_message text,
@@ -321,6 +331,13 @@ create table model_requests (
     provider_observation_json is null
     or jsonb_typeof(provider_observation_json) = 'object'
   ),
+  constraint model_requests_service_tier_ck check (
+    service_tier is null
+    or (
+      octet_length(service_tier) between 1 and 64
+      and service_tier !~ '[[:cntrl:]]'
+    )
+  ),
   constraint model_requests_cost_ck check (
     cost_source in ('provider_reported', 'calculated', 'unavailable')
     and (
@@ -414,10 +431,6 @@ create index model_requests_running_deadline_idx
 create index model_requests_retention_idx
   on model_requests (completed_at)
   where outcome <> 'running';
-create unique index model_requests_client_response_uq
-  on model_requests (client_response_id)
-  where client_response_id is not null;
-
 create table ops_events (
   id text primary key,
   model_request_id text,
diff --git a/backend/migrations/0002_drop_opaque_response_id_index.sql b/backend/migrations/0002_drop_opaque_response_id_index.sql
deleted file mode 100644
index fed242688..000000000
--- a/backend/migrations/0002_drop_opaque_response_id_index.sql
+++ /dev/null
@@ -1,3 +0,0 @@
--- Provider response IDs are opaque protocol values, not relational identifiers.
--- Indexing the raw text makes PostgreSQL reject otherwise valid long handles.
-drop index model_requests_client_response_uq;
diff --git a/backend/migrations/0003_store_opaque_response_ids_as_bytes.sql b/backend/migrations/0003_store_opaque_response_ids_as_bytes.sql
deleted file mode 100644
index ec9716bc4..000000000
--- a/backend/migrations/0003_store_opaque_response_ids_as_bytes.sql
+++ /dev/null
@@ -1,7 +0,0 @@
--- Response IDs are opaque JSON strings. PostgreSQL text rejects NUL even though JSON accepts it.
--- Store their UTF-8 bytes without an index; Rust decodes them only at the audit/API boundary.
-alter table model_requests
-  alter column client_response_id type bytea
-    using convert_to(client_response_id, 'UTF8'),
-  alter column upstream_response_id type bytea
-    using convert_to(upstream_response_id, 'UTF8');
diff --git a/backend/migrations/0004_add_model_request_service_tier.sql b/backend/migrations/0004_add_model_request_service_tier.sql
deleted file mode 100644
index ee1637e4e..000000000
--- a/backend/migrations/0004_add_model_request_service_tier.sql
+++ /dev/null
@@ -1,9 +0,0 @@
-alter table model_requests
-  add column service_tier text,
-  add constraint model_requests_service_tier_ck check (
-    service_tier is null
-    or (
-      octet_length(service_tier) between 1 and 64
-      and service_tier !~ '[[:cntrl:]]'
-    )
-  );
diff --git a/backend/migrations/0005_allow_quota_exhausted_cooldown.sql b/backend/migrations/0005_allow_quota_exhausted_cooldown.sql
deleted file mode 100644
index 4a7f3b94d..000000000
--- a/backend/migrations/0005_allow_quota_exhausted_cooldown.sql
+++ /dev/null
@@ -1,12 +0,0 @@
-alter table provider_accounts
-  drop constraint provider_accounts_cooldown_ck,
-  add constraint provider_accounts_cooldown_ck check (
-    (
-      availability != 'cooldown'
-      or cooldown_until is not null
-    )
-    and (
-      cooldown_until is null
-      or availability in ('cooldown', 'quota_exhausted')
-    )
-  );
diff --git a/backend/migrations/README.md b/backend/migrations/README.md
index 3006980f4..fb912de0e 100644
--- a/backend/migrations/README.md
+++ b/backend/migrations/README.md
@@ -6,14 +6,16 @@ sqlx 在服务启动时（`serve` 监听之前）按编号顺序执行本目录
 
 ## 冻结规则
 
-- 已合入 main 的迁移文件**永久冻结**，一个字节都不能改——包括注释和空白。
-  任何改动都会让所有已部署实例（含自更新链路上的存量用户）在下次重启时
-  因 checksum 失配而无法启动，且没有自动恢复路径。
-- schema 变更一律新增编号迁移（`0003_...`、`0004_...`），永远不回改旧文件。
+- 同一大版本内，已合入 main 的迁移文件**永久冻结**，一个字节都不能改——包括注释和
+  空白。任何改动都会让该大版本的已部署实例在下次重启时因 checksum 失配而无法启动，
+  且没有自动恢复路径。
+- 同一大版本内的 schema 变更一律新增编号迁移（`0002_...`、`0003_...`），永远不回改
+  旧文件。切换大版本时允许把已完成升级的 schema 折叠成新的 `0001` 基线，因为跨大版本
+  不支持在线升级，目标版本必须使用全新数据目录部署。
 - `.frozen-sha256` 是冻结清单（`sha256sum` 格式）。CI 会校验：清单内文件
   字节未变，且目录里每个 `*.sql` 都已入册。新增迁移时在同一提交里执行
   `sha256sum 000N_xxx.sql >> .frozen-sha256` 入册。
-- 若确需修正已冻结迁移里的错误，用新迁移做补偿性变更（`alter` / 回填），
+- 若同一大版本内确需修正已冻结迁移里的错误，用新迁移做补偿性变更（`alter` / 回填），
   不要动原文件。
 
 ## 本地测试库
diff --git a/deploy/README.md b/deploy/README.md
index 2e98a82ea..8931bc036 100644
--- a/deploy/README.md
+++ b/deploy/README.md
@@ -130,10 +130,8 @@ Provider schema 以明文 JSON 保存于 PostgreSQL。Redis 只保存可重建
 ## 镜像升级与源码构建
 
 > [!WARNING]
-> 以下命令只适用于同一大版本内的升级。数据库 schema 与配置只在同一大版本内兼容，
-> 跨大版本直接换镜像会导致服务无法启动；内置一键自更新会拒绝跨大版本目标（3.0 之前
-> 的旧版没有这道门禁，不要用旧版界面的一键更新跨大版本）。跨大版本请全新部署（全新
-> `.runtime/` 数据目录），并重新导入或重新授权 Provider 账号与客户端 Key。
+> 以下命令只适用于同一大版本内的升级，不支持跨大版本在线升级。跨大版本请使用全新的
+> `.runtime/` 数据目录重新部署，并重新导入或重新授权 Provider 账号与客户端 Key。
 
 ```bash
 docker compose -f deploy/compose.yaml build codex-proxy-rs
diff --git a/docs/architecture.md b/docs/architecture.md
index f13590a0c..cdece0ae9 100644
--- a/docs/architecture.md
+++ b/docs/architecture.md
@@ -170,11 +170,11 @@ Admin API 不要求客户端提交配置 revision，也不向客户端暴露配
   客户端可见协议结果。队列不是第二份业务权威，也不做无幂等键的盲目重试。
 - retention 只删除已满足保留规则的历史事实，不改变运行中请求。
 
-业务 schema 由 `backend/migrations/` 下的编号迁移按序定义：`0001_initial.sql` 建表，`0002` 删除
-opaque response ID 唯一索引，`0003` 将 client/upstream response ID 改为无格式假设的 UTF-8 bytes，
-`0004` 增加响应观测得到的 `model_requests.service_tier`。已合入的迁移永久冻结——
-`.frozen-sha256` 是冻结清单，CI 校验清单相对 PR base 只增不改，启动时 sqlx 重校验已应用迁移的
-checksum，不一致直接拒绝启动。规则见 `backend/migrations/README.md`。
+业务 schema 由 `backend/migrations/0001_initial.sql` 直接定义当前大版本的完整新库基线，其中
+response ID 使用无格式假设的 UTF-8 bytes，`model_requests.service_tier` 保存响应观测到的服务档位，
+且额度耗尽账号可以携带冷却截止时间。同一大版本内已合入的迁移永久冻结——
+`.frozen-sha256` 是冻结清单，CI 校验清单相对 PR base 只增不改，启动时 sqlx 重校验已应用
+迁移的 checksum，不一致直接拒绝启动。规则见 `backend/migrations/README.md`。
 
 ## 9. Redis 终态
 
diff --git a/skills/codex-proxy-rs/references/repo-guide.md b/skills/codex-proxy-rs/references/repo-guide.md
index a79d114a7..0e7579a60 100644
--- a/skills/codex-proxy-rs/references/repo-guide.md
+++ b/skills/codex-proxy-rs/references/repo-guide.md
@@ -40,9 +40,9 @@
 
 ## 存储
 
-- PostgreSQL 业务表只有 `0001_initial.sql` 定义的七张；后续编号迁移只做增量。当前 `0002` 删除
-  opaque response ID 索引，`0003` 把 response ID 改为 bytes，`0004` 增加响应观测的 service tier。
-- 已应用迁移按字节冻结：`backend/migrations/.frozen-sha256` 是冻结清单，CI 相对 PR base 做 append-only 校验；schema 变更一律新增编号迁移，规则见 `backend/migrations/README.md`。
+- PostgreSQL 业务表只有 `0001_initial.sql` 定义的七张；该文件直接表达当前大版本的完整新库基线，
+  包括 bytes response ID、响应观测 service tier 与额度耗尽冷却约束。
+- 同一大版本内已应用迁移按字节冻结：`backend/migrations/.frozen-sha256` 是冻结清单，CI 相对 PR base 做 append-only 校验；schema 变更一律新增编号迁移，规则见 `backend/migrations/README.md`。
 - `config_revision` 只用于会改变调度快照或安全配置的管理 mutation。
 - quota、cooldown、catalog generation、自动 refresh 不推进全局 revision。
 - refresh 只推进账号 `credential_revision`；Redis cooldown 是可丢失热缓存。