[Gastown] Dispatch scheduler silently fails all polecats in one rig — different from #1632, #1997, #2260 #4647

What happened?

Confirmed not a duplicate of known issues:

• Not fix(gastown): PR-strategy MR beads stuck in_progress when github_token is missing — agents re-dispatched for completed work #1632 (that's MR-beads stuck in_progress after gt_done — mine fail at dispatch, never reach in_progress)
• Not [Gastown] Mayor control-plane auth remains expired, blocking escalation triage in cloud rig #1997 (Mayor auth is fine — sling/list/update all work)
• Not bug(gastown): create_landing_mr loops infinitely when convoy branches have no PR URL #2260 (that's landing loop after PRs — mine never even produce a PR)

Symptom: Every bead slung to rig ed9475c4-4e70-405b-99dd-c578aa9050d1 fails after 5 dispatch attempts. Applies to seven beads of varying complexity slung over 4 days:

• 5b0b1ba2-9b2e-4302-ac6c-ab9330293aec — pure-text task (read markdown, write markdown). 5/5 failed.
• c5a23ed7-d22f-4bc1-b16b-36772178b5a4 — minimal .docx from scratch. 5/5 failed.
• 79b6a39c-7c16-425d-a62f-b6462a673772 — text redline. 5/5 failed.
• 91ccac29-e322-4fd7-b2b2-61ba841b5c78 — narrow .docx annotation. 5/5 failed.
• 44a49b9b-1509-4eec-b955-d03ce08e986d — python-docx annotation. 5/5 failed.
• e95dff7c-6100-480a-844e-5d9025ea61d8 — write markdown bibliography. 5/5 failed.
• cda85178-01a8-4119-938a-e8cd439511e6 — OOXML tracked-changes. 5/5 failed.

Signature on each failed bead:

• dispatch_attempts: 5
• status: failed
• closed_at: null (never gt_done'd)
• No error message visible on the bead record

Timeline: Same rig had 4 beads succeed on 2026-05-26 (PRs #1, #5, #8, #9 landed). Broke sometime that evening. Nothing since has succeeded.

Additional observation: Mayor's local browse directory /workspace/rigs/ed9475c4-.../repo is now empty (was previously populated with rig files). May be a symptom or an unrelated regression.

Impact: User cannot get any work done in this rig. They have a thesis review workflow that produced valuable output initially but has been stuck for days.

Town: 821c7ac2-7bfe-482e-837f-4ede0bd017df
Rig: ed9475c4-4e70-405b-99dd-c578aa9050d1 (git_url: https://github.com/zannisp/LNU.git)

Area

Agent Dispatch / Scheduling

Context

• Town ID: 821c7ac2-7bfe-482e-837f-4ede0bd017df
• Agent: Mayor (c42850f5-ff49-4281-bf4c-1b94e1a9cab3)
• Rig ID: ed9475c4-4e70-405b-99dd-c578aa9050d1

Recent Errors

No errors visible via the Mayor tools. Would need container startup logs or dispatcher logs to see why polecat processes are failing to start or crashing early in their lifecycle.

Filed automatically by the Mayor via gt_report_bug.