[jbeckwith-oai]

Summary

This pull request fixes queued native tools continuing to run while a selected function-tool failure drains completed output, including after a parallel input guardrail rejects the input.

Selecting a function failure now cancels active native categories and closes their invocation queues. Cleanup preserves the selected error, bounded finalization of completed outputs, genuine caller cancellation, existing approvals, and rejected-input history exclusion. Native error conversion also enters bounded cleanup when cancellation occurred before finalization started. Already-cancelled native tasks receive the existing finalization grace period, then another cancellation if they remain pending; caller cancellation bypasses this grace. Public APIs, configuration, and persisted state are unchanged.

Test plan

• Added streaming and non-streaming regressions for queued native calls, input admission, converted and propagated cleanup errors, caller cancellation, and late finalization timeouts.
• 226 focused tests passed.
• Full verification wrapper passed: formatting, lint, mypy, Pyright, 12,321 parallel tests and 107 serial tests (36 skips across the two runs).
• Local verification used a byte-identical source export under /tmp and a writable XDG_STATE_HOME. The original checkout path contains user, which collides with credential-redaction test fixtures; its default state directory is read-only. All 1,686 tracked files were compared by SHA-256 after verification. No test assertions or source files were changed for this environment adjustment.
• Seven adversarial-review rounds with two independent fresh-context reviewers each, including two rounds for PR feedback. The final two rounds were clean; a subsequent runtime-neutral private typing.cast correction passed the repository type-erasure check and full verification.

Issue number

No public issue.

Checks

• I've added new tests, if relevant
• I've run .agents/skills/code-change-verification/scripts/run.sh
• I've confirmed all verification steps pass
• If using Codex, I've run /review before submitting this PR

[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 2026-10-27T19:11:09.551989Z	54ba662	New commits
🔒 Security Review	✅ Completed 2026-10-27T19:11:49.682477Z	54ba662	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.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: ed8fa0e004

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".