test(lsp): de-flake broker cancel-refresh recovery on slow CI

[ScriptedAlchemy]

broker_cancels_partial_refresh_without_poisoning_warm_client (tests/hooks_lsp_suite/lsp_code_diagnostics_test.rs) is a macOS CI flake. It failed on two consecutive unrelated PRs (#253 perf/git-resolver, #254 refactor/dedup), both with:

panicked at tests/hooks_lsp_suite/lsp_code_diagnostics_test.rs:608:
next refresh should start a clean client and recover:
Config { message: "LSP server '/usr/bin/python3' initialize timed out after 3000 ms" }

Root cause

The test aborts a partial refresh, then asserts the broker isn't poisoned by driving a recovery refresh that must spin up a clean client. That recovery spawns a fresh python3 fake-LSP process; on a loaded runner its cold-start intermittently exceeds the 3s initialize floor (MIN_INITIALIZE_RESPONSE_TIMEOUT). That's a slow start, not a poisoned broker — but the single-shot .expect() treats it as a hard failure. Neither PR touches LSP runtime code.

Fix

Retry the recovery refresh a bounded number of times (5 attempts, 50ms apart), tolerating only the transient initialize timed out error and still panicking immediately on any other error. The EngineState::Ready / total_errors == 1 assertions are unchanged, so the property under test (recovery from an aborted partial refresh) is preserved. No production timeout semantics changed.

Validation

Target test 15/15 green on repeat; full hooks_lsp_suite 116 passed; fmt clean.

🤖 Generated with Claude Code

[changeset-bot]

⚠️ No Changeset found

Latest commit: e14f87e

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.