Port upstream CodexBar 0.56.5 [review]

[Finesssee]

Review status

Review only. Do not merge until explicit approval.

Stacked version-only review delta.

This review branch was rebuilt from current main as a clean stack. Its tree was verified byte-for-byte identical to the already-validated local port tip for 0.56.5, so rebuilding the ancestry did not change implementation content.

Stack

• Head: $(System.Collections.Hashtable.head)
• Base: $(System.Collections.Hashtable.base)
• Next version PR, if any, stacks on this branch.

Validation evidence

Porting work was reviewed with local Codex CLI gpt-5.6-luna:max workers and Thermo-style structural checks. The final 0.56.7 stack is clean and passes:

• Rust cargo fmt --check
• git diff --check
• cargo metadata
• Frontend: 57 files / 328 tests
• TypeScript sc --noEmit
• Vite production build
• Locale parity: 825 keys
• Interaction guard: 9/9

Native Rust test/check execution on the local Windows host is blocked before project linking because the host resolves the wrong GNU/Unix link.exe; this is an environment validation limitation, not a proven semantic porting defect.

Merge policy

Please review this PR and the full stack first. Do not merge yet.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: 0579dcdd-67fd-408a-b603-3b217d0d353b

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[Finesssee]

Thermo-nuclear review: REQUEST CHANGES

1. High - one refresh lifecycle is split across two independent global mutexes. commands/usage_spend.rs adds USAGE_SPEND_SUMMARY_CACHE and a separate USAGE_SPEND_REFRESH_REGISTRY, then manually keeps owner/phase state synchronized with cached-summary refresh_owner. Cleanup mutates the registry and cache in separate lock steps. That is the kind of non-atomic state model the thermo rubric asks us to eliminate: a lock failure or future early-return can leave the two halves disagreeing, and every transition now needs ownership reconciliation. Code-judo: put generation/current phase/cached summary in one UsageSpendCoordinator state behind one mutex, so owner transitions and summary publication are atomic and the manual cross-registry bookkeeping disappears.

2. High - CUA proof is missing. This PR changes TrayPanel, Claude account actions, Codex extra-credit presentation, and the visible Usage & Spend refreshing lifecycle. Repo policy requires a fresh Windows build plus CUA for these flows; none is attached in the PR body/comments.

1k check: no new production threshold crossing.