Fix general automatic recovery of internally paused dispatcher executions

karsten commented

2026-11-11 08:02:35 +00:00

Owner

Fundamental lifecycle fix, not one-off resume of a reported task. Internal runtime-failure pauses/blocks now get persisted automatic due-time recovery just like worker-failed; scheduler adopts diagnosed legacy pauses and resumes SAME execution after backoff and host restart. Sequential retained assessments are independently re-evaluated, not replayed; reviewer-confirmed completion overrides worker self-block without parent decision. Manual pause/cancel/question/deadline/unknown-classifier/policy holds remain binding and clear stale retry ownership. Genuine external holds retain actionable attention alongside scheduled reassessment. Timed internal pauses are routine status, not requests for repeated human nudges. Companion plugin PR #86 pinned immutably. Full pinned suite 1848 total / 1847 passed / 0 failed / 1 skipped; typecheck/build passed; plugins 91/91 passed. Independent diff review findings fixed with regression tests. Upstream settings fix normally merged. No live ledger rewrites, per-task resume or Main fallback. Evidence: development-evidence/automatic-paused-lifecycle-2026-11-11.md.

karsten added 3 commits

2026-11-11 08:02:35 +00:00

fix(dispatcher): automatically recover internally paused executions and retained assessments 222c38dd03

Merge remote-tracking branch 'origin/main' into fix/technical-paused-recovery 5d7bd05276

fix(dispatcher): preserve real hold attention while scheduling internal pause recovery

decision-interface-ci / full-validation (pull_request) Successful in 2m1s

Details

c26d595cbf

karsten added 1 commit

2026-11-11 08:06:10 +00:00

fix(recovery): revalidate scheduled retries and preserve delayed hold attention

decision-interface-ci / full-validation (pull_request) Failing after 1m58s