Flake: an_interrupt_pass_ends_the_whole_build_group fails on macOS ("the pass reaped the build's child itself") #2088

[agent] PR shepherd: the nightly CI schedule run on main (63f791d) failed test (macos-latest, 1) on one core unit test:

utils::process::tests::an_interrupt_pass_ends_the_whole_build_group ... FAILED
panicked at crates/socket-patch-core/src/utils/process.rs:1223:9:
the pass reaped the build's child itself
test result: FAILED. 7736 passed; 1 failed

• Failing run: https://github.com/SocketDev/socket-patch/actions/runs/43682219537/job/132715428064 (2026-11-15 schedule)
• The same commit passed this job on its push run (43556647706) and on the 2026-11-14 schedule run (43562815783), and process.rs hasn't changed since Fix the defects found by the 5.1.0 full review: do no harm, Bundler hosted mode, VEX, monorepo scans #2052. So this is probably a timing flake in the test, not a regression.
• The assertion at process.rs:1221-1224 expects child.try_wait() to return Err (the child was already reaped). On macOS, try_wait seems to return Ok(..) sometimes, so the test races with whatever reaps the child, or with how macOS reports a reaped zombie.

The other red job in that run (test (windows-latest, 1)) is #2080, fixed by #2082.

Expected fix: make the assertion robust to how macOS reports a reaped child. Don't skip or weaken what the test checks: that the interrupt pass ends the whole group without reaping the build's direct child.

Generated by Claude Code