Measured 2026-09-05 08:50Z on the maintainer's Windows box while diagnosing a low-memory kill of two background CI monitors: 33 git fsmonitor--daemon run --detach --ipc-threads=8 processes alive, ages 233 to 4697 minutes, about 137 MB total, every parent gone. Global core.fsmonitor is unset, so the daemons come from repositories that set it locally; the worktree-helper suite's scenario 28 (fsmonitor-hidden-git-add-timeout in scripts/git/Test-New-CodexIssueWorktree.ps1) enables fsmonitor in a fixture repository, and a detached daemon outlives the fixture directory it watched. Cause unverified beyond that correlation; the daemons were not the memory culprit (node and Rider were) and none were killed. Ask: confirm which repositories spawn them (git fsmonitor--daemon status in the fixture during the scenario, and core.fsmonitor in any local repo config), and have the suite stop the daemon (git fsmonitor--daemon stop) before removing a fixture that enabled it, plus a note in ~/.claude/MACHINE.md RAM hygiene if the sweep script should reap aged daemons. Refs #2425, #2670.
Measured 2026-09-05 08:50Z on the maintainer's Windows box while diagnosing a low-memory kill of two background CI monitors: 33
git fsmonitor--daemon run --detach --ipc-threads=8processes alive, ages 233 to 4697 minutes, about 137 MB total, every parent gone. Globalcore.fsmonitoris unset, so the daemons come from repositories that set it locally; the worktree-helper suite's scenario 28 (fsmonitor-hidden-git-add-timeoutinscripts/git/Test-New-CodexIssueWorktree.ps1) enables fsmonitor in a fixture repository, and a detached daemon outlives the fixture directory it watched. Cause unverified beyond that correlation; the daemons were not the memory culprit (node and Rider were) and none were killed. Ask: confirm which repositories spawn them (git fsmonitor--daemon statusin the fixture during the scenario, andcore.fsmonitorin any local repo config), and have the suite stop the daemon (git fsmonitor--daemon stop) before removing a fixture that enabled it, plus a note in~/.claude/MACHINE.mdRAM hygiene if the sweep script should reap aged daemons. Refs #2425, #2670.