test: align eval host startup and cleanup deadlines

[vriesd]

Summary

• Give the real-host integration test a 240-second envelope around the existing 180-second startup contract, retaining the actual native tuple assertion.
• Route startup timeout through cancellation and preserve environment/scratch cleanup.
• Cover cancellation during a pending readiness POST with credential-preservation assertions.

Validation

• Independent Flow final review passed for the bounded repair.
• Fresh pinned Linux, no credentials and no warm-up: 821 passed, 1 skipped; host test completed in 151 seconds, exceeding its former 30-second limit.
• Linux replay 13/13, audit and live smoke passed.
• Focused host/reporting tests: 106 passed.
• Clean-worktree contribution preflight: 821 passed, 1 skipped; typecheck, lint, metadata and build passed.
• Canonical grader remains unchanged by this test-only repair. No runtime, package, version or qualification-policy changes.
• Optional gitleaks unavailable. Historical campaign/canary files and local diagnostic helpers are excluded.

This PR does not claim release qualification or authorize publication.

[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 2026-09-06T00:27:01.777825Z	6d32cad	PR opened

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: 6d32cad61f

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".

[chatgpt-codex-connector]

Preserve the host's full startup window

On a slow filesystem, this signal can cancel an otherwise valid startup early: its 180-second clock begins when EvalHost.start is invoked, but the harness does not begin its own STARTUP_TIMEOUT_MS deadline until after credential copying, fixture creation, synchronous Git commands, package-cache copying, and process spawning. Consequently, that setup time is subtracted from the host's intended 180-second health/readiness allowance, making this integration test flaky in environments where startup legitimately approaches the internal deadline; give the cancellation signal additional headroom while keeping it below the 240-second outer timeout.

Useful? React with 👍 / 👎.