Check service liveness while monitoring newly started native runs, matching the recovery protection already present during adoption. After three consecutive dead-service probes, reconcile completed results, requeue in-flight items and release the run instead of polling a stale running snapshot indefinitely.
Cover stale snapshots, transient probe failures, paused workers, run isolation, startup timeout and sequential verification reconciliation. Related queue and verification validation: 155 tests passed; scoped analyzer and formatter clean.
Refs #675
zarzet
09e07671c7