test/crew.test.ts runs under 10 s: its archive test waits for the crew, not a fixed 6 s #9089

bun test test/crew.test.ts now runs in 2.86 s, down from 21.49 s, measured on this branch before and after the change with the same 4 tests passing. Only test/crew.test.ts changes.

• The fixed 6 s sleep is gone. In the archive test ("a run queued before Archive is cancelled…"), Wake is now followed straight away by a new run on the same issue, and the test waits for that run to finish ok. Only one run per issue runs at a time, and the cancelled run was queued first, so the crew's next pick after Wake would reach it first if it had been revived. Once the new run finishes, the old one is checked again: still cancelled, its summary is still ARCHIVED_WHY, and its log is still the single archived line. That is more than the old test checked, which was the state alone.
• The crew ticks faster in this file only. It now uses the existing tickMs option (startStandInCrew({ tickMs: 50 })), as crew-declines, crew-pause and landing already do. The production default of 5 s is unchanged. The ticks were the rest of the cost: every wait for a run to finish could sit up to 5 s.
• The test still catches Wake reviving a run. I broke it once on purpose: moveArchived set runs cancelled with ARCHIVED_WHY back to queued on Wake. The test failed: the revived run was in the way, so the new run on that issue was refused. Afterwards I restored the code.
• Per-test times: the whole file takes 2.86 s, so every test in it is far under 5 s. Bun doesn't print each test's time when they pass, so there's no per-test figure to quote.

bun scripts/gate.ts and bun x tsc --noEmit pass. I couldn't update the plan on the issue because the keep_plan tool wasn't available in this run.

To try it: bun test test/crew.test.ts

🤖 Generated with Claude Code

Closes #9086

bun test test/crew.test.ts now runs in 2.86 s, down from 21.49 s, measured on this branch before and after the change with the same 4 tests passing. Only test/crew.test.ts changes.

• The fixed 6 s sleep is gone. In the archive test ("a run queued before Archive is cancelled…"), Wake is now followed straight away by a new run on the same issue, and the test waits for that run to finish ok. Only one run per issue runs at a time, and the cancelled run was queued first, so the crew's next pick after Wake would reach it first if it had been revived. Once the new run finishes, the old one is checked again: still cancelled, its summary is still ARCHIVED_WHY, and its log is still the single archived line. That is more than the old test checked, which was the state alone.
• The crew ticks faster in this file only. It now uses the existing tickMs option (startStandInCrew({ tickMs: 50 })), as crew-declines, crew-pause and landing already do. The production default of 5 s is unchanged. The ticks were the rest of the cost: every wait for a run to finish could sit up to 5 s.
• The test still catches Wake reviving a run. I broke it once on purpose: moveArchived set runs cancelled with ARCHIVED_WHY back to queued on Wake. The test failed: the revived run was in the way, so the new run on that issue was refused. Afterwards I restored the code.
• Per-test times: the whole file takes 2.86 s, so every test in it is far under 5 s. Bun doesn't print each test's time when they pass, so there's no per-test figure to quote.

bun scripts/gate.ts and bun x tsc --noEmit pass. I couldn't update the plan on the issue because the keep_plan tool wasn't available in this run.

To try it: bun test test/crew.test.ts

🤖 Generated with Claude Code