fix(integrations): bind connector reads to request lifetimes

[Chris0Jeky]

Closes #3327.

Summary

Binds integration list/detail reads to explicit request lifetimes instead of entity ID or one shared Boolean:

• list and detail use independent ownership lanes;
• a newer read retires only the previous owner in the same lane;
• reset advances an epoch before clearing visible state;
• stale successes, failures, toasts and finally settlement cannot alter the replacement lifetime;
• loading remains true while either current list or detail lane still owns work;
• legitimate list/detail concurrency remains available.

The integration API, DTOs, routes and mutation behavior are unchanged. This slice does not claim to cancel or roll back a connector mutation already accepted before reset.

Test-first evidence

Base: 307c3b8b50bec1cb0bfaea3e570a942bcb1d4451

• Test-only commit: 78ecac7d3bc37fe80ba1c746f0eecb9e6d72f372
• Corrective exact head: 4e232e22778489681b9db5746e5c221ada7a088a

A supplemental runner transpiled and executed the actual production store with only its framework/API boundaries stubbed:

• unchanged main: 0/5 ownership schedules passed;
• corrected source: 5/5 passed.

The committed Pinia/Vitest suite covers six schedules:

1. A-old → B → A-new with A-old settling last;
2. reset followed by same-ID stale success;
3. reset followed by same-ID stale failure and no stale toast;
4. list response after reset;
5. reverse-settling list reads;
6. overlapping list/detail loading ownership.

The test-only Required CI was superseded and cancelled when the corrective head was pushed, so no hosted canonical-red claim is made. Smart CI and Extended passed on the test-only commit; canonical regression qualification belongs to the corrective exact head.

Verification

Local/source-backed:

• combined integration/permission actual-module ownership suite: 12/12 passed after correction;
• TypeScript syntax transpilation of changed store/tests passed;
• docs governance and golden-principles checks passed;
• relative-link scan checked 719 Markdown files with zero broken links and the same three pre-existing masking warnings.

Hosted exact-head:

• Smart CI Self-Test 35597534968: success;
• CI Extended 35597535200: success;
• Required CI 35597535219: success across the complete required matrix.

Ready for independent review. Green CI is qualification evidence, not merge authorization. Review should focus on A→B→A identity reuse, reset ownership, stale error/toast suppression, and independent list/detail loading.

No merge, release or deployment qualification is claimed.

[Chris0Jeky]

Fresh source-backed verification for exact head 4e232e22778489681b9db5746e5c221ada7a088a:

• Actual production modules, unchanged uploaded-main source: 0/12 combined integration/permission ownership schedules passed; every failure matched the named stale-state or loading symptom.
• Actual production modules, corrected source: 12/12 passed.
• integrationStore.ts, integrationStoreOwnership.spec.ts, permissionsStore.ts, permissionsStoreOwnership.spec.ts, and adjusted permissionsStore.spec.ts all transpile under TypeScript 5.8.3 with zero diagnostics.
• Smart CI Self-Test 35597534968: success.
• CI Extended 35597535200: success.
• Required CI 35597535219: queued at this checkpoint.

The actual-module runner uses only framework/API boundary stubs and does not replace canonical Pinia/Vitest or full hosted qualification. The test-only Required run was superseded/cancelled, so this comment does not claim a hosted canonical-red execution.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-21T14:09:04.107638Z	4e232e2	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[Chris0Jeky]

Fresh-context continuation review: read the complete three-file diff and the unchanged surrounding mutation paths. Unique request tokens plus reset epoch prevent A-to-B-to-A identity reuse; list and detail retain independent ownership, and current loading derives from both owners. Canonical deferred tests assert the installed result, stale failure/toast suppression, and loading while the other lane remains pending. No blocking finding in this read-only slice.

Current-main integrationStore.ts blob 6bb30b2073b1879b1fde9679f7828c2c43a9e3c7 matches the inspected unchanged base; no overlapping source rewrite is being assumed away. Exact-head Required CI 35597535219, Self-Test 35597534968 and both returned Extended runs succeeded. Codex completed at this exact head, and the current inline thread list is empty.

The real session lifecycle/mutation admission follow-up remains in existing child #3332, followed by #3336 for same-connector mutation ordering. This parent does not claim those remaining guarantees, and those children must be retargeted/requalified after it lands. Normal expected-head merge under the maintainer's current instruction; no new local Pinia/Vitest execution or release qualification is claimed.