[Gastown] Dispatch/reconciler loop appears stopped — all agents idle, no beads dispatched for 20+ min despite interventions

[kilo-code-bot]

What happened?

In town 8bc2f148-811a-48f2-a730-e13029edb1fe, rig 64b59606-4597-4237-a56a-fb9bde8f96e7, the dispatch/reconciler loop appears to have stopped running. All 19 agents (18 polecats + 1 refinery) are idle with current_hook_bead_id=null and last_activity_at frozen at ~2026-07-07T00:28:38Z. No beads have been dispatched for 20+ minutes despite a full backlog.

EVIDENCE:

1. 12+ unblocked open issue beads (create-story 1.1, 1.2, 1.3, 1.6, 1.7, 2.4, 3.2, 4.2, 5.1-5.4) are permanently backed off at dispatch_attempts=4, last_dispatch_attempt_at=2026-07-07T00:23:04Z. None have been re-dispatched.
2. 5 story beads are in_review status (1.5, 2.2, 2.3, 3.1, 4.1) with open PRs (feat: enable strict tool use for Anthropic models #80-83 on StirlingGoetz/ForgetfulDrinker). PR feat: enable strict tool use for Anthropic models #80 shows merge_state_status=CLEAN, awaiting_approval=0.
3. 5 open merge_request beads are assigned to the refinery (assignee_agent_bead_id = refinery agent d725aa5c) BUT the refinery's current_hook_bead_id=null — it is not being hooked to its assigned merge_requests.
4. One validate bead (validate-story 1.4) has its dependency satisfied (create 1.4 is closed) yet remains undispatched (dispatch_attempts=0).

INTERVENTIONS ATTEMPTED (all no-ops):

• gt_nudge to refinery (mode=immediate) — queued, refinery stayed idle
• gt_agent_reset on 3 idle polecats (Toast, Maple, Birch) — reset to idle, not re-hooked
• gt_bead_update priority->high on 12 open create-story beads — dispatch_attempts unchanged, no re-dispatch
• gt_agent_reset on refinery itself — reset to idle, not re-hooked to merge_requests
• Waited ~3 min total across checks — zero state change to any agent or bead

CONVOYS AFFECTED:

• 775d9096 (implement-story Epic 1): 14 beads, 1 closed, 1 in_review
• 27e143af (implement-story Epics 2-5): 24 beads, 1 closed, 4 in_review
  Both merge_mode=review-then-land, status=active, not landed.

The dispatch scheduler is not cycling. Agent resets and priority bumps do not re-trigger dispatch. This looks like a stuck/crashed reconciler rather than a transient backoff.

Area

Agent Dispatch / Scheduling

Context

• Town ID: 8bc2f148-811a-48f2-a730-e13029edb1fe
• Agent: Mayor (206d613e-2797-479b-a8bb-893a6db17095)
• Rig ID: 64b59606-4597-4237-a56a-fb9bde8f96e7

Recent Errors

No explicit error messages returned by tool calls — all tool calls succeeded. The failure is silent: no agent transitions from idle, no bead gets dispatched or hooked despite valid open work and assigned merge_requests. gt_list_rigs also returned 'No rigs configured' on first call despite the rig existing (returned empty), which may indicate a broader control-plane flakiness.

Filed automatically by the Mayor via gt_report_bug.