Eval queue depth over time

Pending issues (queued for evaluation), sampled from worker polls of /api/eval/next. Only accumulates while some worker (in-cluster or remote) is actively polling.

Daily invalidations by signal

Rolling 7-day total: 0 invalidations.

Issues invalidated (evidence_generation bumped) per day by the commit scanner, broken out by which signal fired: an exact cross-repo reference, a changed evidence path, a semantic commit match, a same-repo bare #N reference, or a cross-source LP: #N Launchpad reference. Note: an issue matched by two signals in one pass is counted in both buckets but bumped once, so the stacked total overstates unique invalidations — the rolling total above is the honest headline.