test(extension-host): kill the host under a command with the default deadline

[Lstarsky0]

What and why

extension_host::tests::a_slow_command_is_cancelled_and_a_killed_host_fails_it_as_down has failed twice on windows-latest since 10-10, once on main:

• https://github.com/codewhale-hq/Codewhale/actions/runs/38021394688/job/114123124425 (the run for a dependabot bump, f50bc9e)
• https://github.com/codewhale-hq/Codewhale/actions/runs/38123604545/job/114423296927 (main, 25cc4f7)

timed out after 0s and was cancelled

The test runs both of its halves on one manager whose command_run_deadline is 300 ms. The first half needs that: it checks the deadline cancels a slow command. The second half starts a slow command, sleeps 100 ms, kills the host with taskkill and expects extension host is down: exited:. That leaves the kill 200 ms to land. When it takes longer, the command ends on its own deadline first, which is the message above.

The kill half now runs on a second manager from supervised_manager, which keeps the default 30 s command deadline, so the only thing that ends the command there is the host going away. The tool-side tests are already split this way (a_call_past_its_method_deadline_is_cancelled_and_the_host_stays_usable and killed_host_fails_calls_once_and_replays_with_fresh_owners). The first half is unchanged.

The comment about the kill landing inside the 300 ms deadline no longer describes anything, so it is removed.

Issue

No-Issue: Windows CI flake, no issue filed for it

How I tested it

• To make the kill slow on purpose I made one scratch build that keeps the old test next to the new one and reads the sleep before the kill from an environment variable. At the normal 100 ms both pass 5 of 5. At 400 ms and at 2 s the old test fails 5 of 5 with the message above and the new one passes 5 of 5. None of that scratch code is in the diff.
• cargo fmt --all -- --check, cargo clippy -p codewhale-tui --all-targets --all-features with CI's flags, and cargo check -p codewhale-tui --all-targets: clean. Clippy also on 1.99, plus cargo +1.89.0 check -p codewhale-tui.
• cargo nextest run -p codewhale-tui --lib: 15286 passed, none failed.

No new tests; this changes one existing test.

Checklist

• One focused change, rebased on current main
• cargo fmt --all passes, and tests cover new or changed behavior
• If this adds or changes a user-facing feature, I updated its docs and its row in docs/features.toml (cargo test -p codewhale-tui --test feature_registry checks it) — n/a

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[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 2026-10-14T05:30:07.646748Z	3411127	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.

[Hmbown]

Thank you, @Lstarsky0, for separating the explicit 300 ms cancellation check from the killed-host check under the default 30 s command deadline. That preserves the host-down, restart identity, stale-generation refusal and fresh-echo assertions while removing the artificial deadline competition in the second half. Your PR has merged as its own work in 88b4c58dfb; the source branch is retained.

Exact PR-head CI completed with macOS 19,038 passed / 0 failed / 33 skipped, Ubuntu 19,035 / 0 / 34, Windows 18,304 passed (one existing browser-launcher leak) / 0 failed / 27 skipped, plus CI-flags Clippy and all ratchets. The changed fixture and owner-crash recovery passed on each native platform. This is verification of the PR source, not final installed-package or VM acceptance. It is carried toward 0.10.2, which has not shipped yet. Thank you for making the Windows coverage more useful without weakening the original cancellation assertions.