[tianxiaochannel-oss88]

The workspace handoff regression assumes that an HTTP request has entered the server mutation queue after a 50 ms sleep. Under CI load, the later reply can overtake the request and fail the final binding assertions. This failed in the performance candidate run and again in the post-merge scheduled run 40855279750.

Wait for the server to finish reading the handoff request body and cross the handler's microtask boundary before submitting the later reply. Keep every binding and factory-call assertion, and bound the test with the existing 10-second async timeout. No product code changes.

The full-ci label helper also fails with fatal: not a git repository because it intentionally has no checkout. Pass the repository explicitly to gh workflow run, retaining the fork guard, permissions and checkout-free execution.

Validation

• Artificially delaying request-body consumption by 150 ms reproduces the handoff failure before the change; the corrected test passes with the same delay.
• Removing mutation serialization makes the corrected test fail on the final binding, confirming it still detects a real coordination regression. The generated module was restored after this probe.
• Complete local channel-workspace.test.mjs: 226 passed, 0 failures.
• Node 22.22.1 CI contract checks after integrating current main: 221 passed, 1 environment-dependent skip, 0 failures.
• Both workflow YAML files parse; git diff --check passes.

Full target-runner CI 40918754725 passed on commit 18c425855793119719f93afb33cbe8df5d470e20: all four real macOS daemon shards, unit/Web tests, type checks, dependency/security checks, Linux contracts, native helper smoke and the final quality gate succeeded. The handoff regression passed in shard 0.

Daemon test steps took 17m44s, 14m24s, 12m33s and 11m10s for shards 0 through 3. The complete run took 23m57s, including setup and scheduling. These are observed durations, not a controlled comparison of total CI speed.

Follow-up to the merged performance change #618. The branch includes current main 7c7861020fc3eacddd284459c4d6005c74a554f6. The final head 4c22eff501e890c084fa0a2dca7cc7358a7bc7ad retains the same three-file change. Local CI contract checks again passed (221 passed, one environment skip).

Final pull-request CI 40994387706 passed on that exact head, together with CodeQL: all 15 checks succeeded. The full CI run took 24m29s; daemon test steps for shards 0 through 3 took 13m46s, 10m04s, 8m12s and 9m33s, and unit tests took 18m13s. The handoff regression passed in shard 0. These are observed durations, not a controlled speed comparison.

The label helper runs from the base branch, so its changed workflow itself cannot execute on a label event until integrated; the exact explicit-repository dispatch command was successfully exercised from a directory without a Git repository. No product behavior or deployment changes are included.

[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 2026-10-30T02:48:54.328060Z	87931ec	Draft marked ready
🔒 Security Review	✅ Completed 2026-10-30T03:01:57.024042Z	4c22eff	New commits

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.