The `script run command > runs a script and returns result` test runs a
trivial, deterministic bun script and asserts exit code 0. On CI it
intermittently fails when the standalone worker (notably on Windows)
transiently fails to execute the job — identical bun jobs complete
successfully elsewhere in the same backend session, so the failure is
environmental, not a regression.
Two problems made this both flaky and undiagnosable:
- `--silent` plus asserting only on `result.code` meant the job's actual
error never reached the CI log, so a flake left no trace.
- No test-level retry, so a single transient worker hiccup failed the run.
Add `retry: 2` to the two worker-executing tests in the block, and
include stdout/stderr in the assertion label so the next occurrence is
debuggable.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Ruben Fiszel
3d6e8b1153