test/crew-review.test.ts runs under 10 s #3650

test/crew-review.test.ts now runs in about 1.2 s, down from 10.8 s, and scripts/test-times.json records 1.2. I ran it ten times in a row and all ten passed (1.15–1.30 s each). I did not get a type check: bun x tsc --noEmit failed on a read-only temp directory and there is no local tsc binary. The forge's check will run it on this push.

Cause. The run itself takes about 50 ms. The time was the crew dispatcher's 5 s poll (TICK_MS in src/crew/dispatch.ts). Each queued run waited for the next tick, and the test queues two runs. No production code changed; startCrew already takes a tickMs option.

Change.

• startStandInCrew in test/helpers/crew-standin.ts takes an optional { tickMs } and passes it to startCrew. Without it, behaviour is unchanged for the other crew*.test.ts files.
• test/crew-review.test.ts boots with tickMs: 20.

Coverage. No test was deleted, and nothing the file asserts was weakened. The same single test still checks the verdict, the review state, no comments, the second pass without --resume, and no duplicate review.

Not done.

• The fake harness is still the shell-script stand-in from the shared helper, and the helper's agentRunSettled still polls every 100 ms. I did not swap in a fake clock, because that would mean changing the shared helper and production timers. The file was over budget only because of the dispatcher tick.
• The file has one test, so "every test under 1 s" holds only roughly. The forge boots once per file, in beforeAll, at about 240 ms. The test body takes under a second but I didn't measure it exactly; the verbose run I tried it with didn't print per-test times.

Try it. bun test test/crew-review.test.ts

🤖 Generated with Claude Code

Closes #3594

test/crew-review.test.ts now runs in about 1.2 s, down from 10.8 s, and scripts/test-times.json records 1.2. I ran it ten times in a row and all ten passed (1.15–1.30 s each). I did not get a type check: bun x tsc --noEmit failed on a read-only temp directory and there is no local tsc binary. The forge's check will run it on this push.

Cause. The run itself takes about 50 ms. The time was the crew dispatcher's 5 s poll (TICK_MS in src/crew/dispatch.ts). Each queued run waited for the next tick, and the test queues two runs. No production code changed; startCrew already takes a tickMs option.

Change.

• startStandInCrew in test/helpers/crew-standin.ts takes an optional { tickMs } and passes it to startCrew. Without it, behaviour is unchanged for the other crew*.test.ts files.
• test/crew-review.test.ts boots with tickMs: 20.

Coverage. No test was deleted, and nothing the file asserts was weakened. The same single test still checks the verdict, the review state, no comments, the second pass without --resume, and no duplicate review.

Not done.

• The fake harness is still the shell-script stand-in from the shared helper, and the helper's agentRunSettled still polls every 100 ms. I did not swap in a fake clock, because that would mean changing the shared helper and production timers. The file was over budget only because of the dispatcher tick.
• The file has one test, so "every test under 1 s" holds only roughly. The forge boots once per file, in beforeAll, at about 240 ms. The test body takes under a second but I didn't measure it exactly; the verbose run I tried it with didn't print per-test times.

Try it. bun test test/crew-review.test.ts

🤖 Generated with Claude Code

I merged main into the branch and pushed it. The only conflict was in test/helpers/crew-standin.ts.

• Conflict: both sides added a tickMs option to the crew stand-in's setup. I kept main's tickMs: options.tickMs line and dropped this branch's conditional-spread version. After the merge, that file is identical to main's.
• Other files: the scripts/test-times.json merge needed no changes.
• Checks: test/crew-review.test.ts passes in about 1.3 s. The type check printed nothing visible, because Bun errored with unable to write files to tempdir: ReadOnlyFileSystem. The type check may not have actually run. I did not run the full bun test, since the forge runs the whole check on the push.

To try it, run bun test test/crew-review.test.ts.

I merged main into the branch and pushed it. The only conflict was scripts/test-times.json, and I kept this branch's timings for the two crew-review entries.

• Resolution: I kept this branch's crew-review-budget (15.8 s) and crew-review (1.2 s). The change here is splitting test/crew-review.test.ts so it runs in under 10 s, and main's numbers (1.5 s and 11.6 s) describe the file before that split. All other timings from both sides merged cleanly.
• Tests: test/shards.test.ts and test/crew-review* gave 70 passes and 1 failure. The failure is the check in test/shards.test.ts that needs typescript/package.json, which isn't installed in this sandbox. I didn't run bun install or touch that test.
• Not run: I didn't run the type check, because I never installed dependencies (bun install). The forge's check on this push will cover it.

I restored test/crew-review-budget.test.ts to 1.5 in scripts/test-times.json, matching main, and pushed. I didn't run any tests for this one-value change.

You were right that 15.8 was a stale value from my merge resolution. The only difference from main in that file is now test/crew-review.test.ts, which goes from 11.6 to 1.2.

The rest of this pull is unchanged: tickMs: 20 on the stand-in crew in test/crew-review.test.ts, which removes the 5 s dispatcher wait.