While shepherding kompile-core PR 217 through the merge queue on 2026-07-14, build-watchman 0.0.12

Reported (UTC): 2026-07-14 15:22

While shepherding kompile-core PR 217 through the merge queue on 2026-07-14, build-watchman 0.0.12 was relaunched immediately after a manually diagnosed second queue eviction and successful GraphQL enqueuePullRequest. GitHub had already created visible queue SHA 811181186dada90f2936948373b9dde259e0e6a2 with all three checks dispatched, but watchman attempted enqueue again; GitHub returned 'Pull request is already in the queue', which watchman treated as terminal PROBLEM and exited, then misleadingly reported head check-suite failures from prior attempts. Impact: the required --to-merged monitor silently stopped even though the queue entry was healthy, risking an unwatched merge. Workaround: verify the queue ref and check runs directly, then launch watchman a second time after the queue ref is visible; the second process attached correctly. Suggested durable fix: treat enqueuePullRequest 'already in the queue' as an idempotent success, refresh mergeQueueEntry/queue ref, attach to it, and ignore superseded prior head/queue check attempts.

Production verification — 2026-07-17

Status: STILL EXISTS. Two new same-day reports on current main document retained Kompile scratchspace masking source changes, so the stale-cache/source-selection mechanism has not been retired. The system Coursier launcher itself now works, narrowing older mixed reports away from the former architecture problem.

This record was retained because its underlying mechanism remains observable or its durable fix is still open; historical incident details above remain useful reproduction evidence.