Windows CI: comment-checker downloader test cold-starts six probe processes and one hits its 10 s abort (exit 143)

[code-yeongyu]

Summary

packages/omo-opencode/src/hooks/comment-checker/downloader.test.ts failed on test (windows-latest, 1/2) (run 36323638339): "comment-checker release download > uses the pinned release without npm resolution on linux/arm64" got exitCode: 143 after 10.6 s.

Root cause

The test spawned one cold bun run of the probe fixture per target platform: six full import-graph cold starts, each under AbortSignal.timeout(10_000). On a loaded Windows runner one cold start exceeded 10 s, the abort sent SIGTERM (143), and the test failed. Nothing about the linux/arm64 target was wrong. The downloader reads process.platform/process.arch at call time, so the six spawns were never needed.

Expected

• One probe process covers every target, so there is one cold start instead of six.
• Every per-target assertion is unchanged: the pinned release URL, zero npm resolution attempts, and a graceful null result. There is also one assertion that the probe exits cleanly.
• No timeout is raised.

Refs #8324

[code-yeongyu]

Fixed by #8961 (merge commit 6f65ada): every download target is probed in one process instead of six cold Bun children.

Evidence: 3 focused windows-latest soaks green before merge (in the PR); the combined fix set was full-matrix green 3 times on the integration head; after merge, dev push CI run 36360065018 on e89647d (which contains every win-ci fix) was green in all 3 attempts on every Windows job, with the test steps actually executed.