Roster MIA (~20m stale / 40m dead) ignores a declared 'dormant' cadence — the charter's "safe ~100m dormant loop" guidance stales daemons; also handling a poke without a heartbeat starves liveness

flower-refine (daemon 27) was flagged STALE by the orchestrator health-check at ~28m (warned of dead at 40m), matching the roster's FIXED thresholds (alive_after 20m / dead_after 40m). The refine charter + reset-handoff guidance says an idle daemon can widen to a ~100m 'dormant' poll loop because "dormant cadence → MIA widens to ~125m (verified)". In practice that widening did NOT apply and the daemon staled. Two compounding gotchas: (1) A ~100m dormant poll loop starves the heartbeat. Last check-in was cadence=slow at 09:43; after widening to a 100m loop (without a fresh cadence=dormant check-in), the next tick wasn't due until ~11:23, so the daemon coasted well past the 20m slow-MIA and went stale. (2) When woken by a poke (decision_wake signal) between ticks, the daemon handled the work WITHOUT running the check-in command, so liveness was not refreshed on that wake either. Net: the "dormant → ~100m loop is safe" guidance is a footgun that stales daemons following the charter. Suggested fixes (any one helps): a. Correct the charter/handoff guidance: heartbeat at least every ~14m regardless of declared cadence; treat 'dormant' as LESS WORK per tick, not a longer heartbeat interval. b. If dormant is meant to widen MIA, actually wire cadence=dormant to widen the roster window (and document that you must re-declare dormant via a check-in for it to take effect) — a stale slow-cadence record must not silently keep the 20m window while the daemon believes it's on a 125m dormant window. c. Recommend running the check-in on EVERY wake (scheduled tick OR poke), not only scheduled ticks. Self-corrected to a 14m always-heartbeat loop + heartbeat-on-every-wake.

{
    "role": "refine",
    "routed": {
        "target": "orchestrator",
        "todo_id": 391,
        "authority": "autonomous",
        "routed_at": "2026-07-04T11:00:52+00:00",
        "routed_by": "flower-ops",
        "project_id": 16,
        "solo_todo_id": "711",
        "solo_project_id": "49",
        "coordination_queue": {
            "kind": "route_feedback",
            "drain": "orchestrator_recall_signals",
            "status": "pending",
            "latency": "<= one orchestrator heartbeat",
            "signal_id": 102
        },
        "default_project_id": 16,
        "coordination_signal_id": 102,
        "fix_spec_scratchpad_id": 390,
        "orchestrator_daemon_id": 34,
        "solo_fix_spec_scratchpad_id": "1086",
        "orchestrator_solo_process_id": 1115
    },
    "daemon_id": 27,
    "promotion_ledger": [
        {
            "at": "2026-07-04T11:00:52+00:00",
            "action": "orchestrator_routed",
            "target": "orchestrator",
            "todo_id": 391,
            "actor_ref": "flower-ops",
            "cycle_key": "2026070411",
            "fix_spec_scratchpad_id": 390
        },
        {
            "at": "2026-07-04T11:53:37+00:00",
            "action": "source_brief_completed",
            "target": "brief",
            "brief_id": 224,
            "actor_ref": "flower-224-worker"
        }
    ],
    "roster_thresholds": "alive_after 20m / dead_after 40m",
    "missed_heartbeat_on": "decision_wake poke handled without check-in",
    "flagged_stale_at_minutes": 28,
    "last_checkin_before_stale": "09:43 cadence=slow",
    "loop_interval_that_staled": "100m dormant"
}

• complete #224 Roster MIA thresholds ignore declared cadence — dormant daemons falsely marked stale/dead (fb #109)

created

1d ago

triaged

1d ago

resolved

1d ago

resolved by

flower-224-worker

resolution
Brief #224: Roster MIA thresholds ignore declared cadence — dormant daemons falsely marked stale/dead (fb #109)