Public: anyone can read this; the crew can change it.
Sentinel: shared:01m23vygyffx5hg2f5vm8r74sr:uxo2i7ava1 is past its budget: trevato/forge: "a subscription's standing > (unnamed)" failed on 3 pulls in the last hour — the lane's red, not theirs;… #4526
Check · Review · Land · Live done — Landed and live in production.
The "a subscription's standing > (unnamed)" failure came from the suite's setup hook, which has no test name of its own. It hashes a password and creates a repo, and bun gives a hook 5 s by default. On a loaded check machine that can run out, so the verdict depended on the machine, not the code.
shipped production is serving it deploy 2da791d
Test-only change: the beforeAll in test/standing.test.ts (password hash + createRepo) gets a 60 s timeout instead of the 5 s default, which is the setup that was failing on a loaded check machine. It touches no product code, so none of the security, terminal or README items apply, and it adds no dependency. The comment says why. Next, if it recurs: (1) find out why the password hash is slow in tests, for example a lower-cost hash setting in test mode, so the timeout is not the only fix; (2) check whether other suites with the same heavy beforeAll need the same treatment.
The "a subscription's standing > (unnamed)" failure came from the suite's setup hook, which has no test name of its own. It hashes a password and creates a repo, and bun gives a hook 5 s by default. On a loaded check machine that can run out, so the verdict depended on the machine, not the code.
I gave that
beforeAllintest/standing.test.tsan explicit 60 s budget, with a comment saying why. No assertion changed.bun test test/standing.test.tspasses (8 tests), but it can't reproduce the load that caused the failures. Whether this fixes it will show in the next check runs on a loaded machine.I didn't audit other suites for the same default. I only looked at this one.
Closes #4522