Diagnose ~/bin/gh-fleet-token.sh (RE-QUEUED for current >12h GH auth outage)

Fresh re-queue of the 5/25 task (id 6651d747...) which completed within minutes and cannot re-trigger automatically. Current outage: GH auth continuously down since ~2026-05-26 00:30Z (~28h at 5/27 04:20Z reflection). Steps: 1) Without set -x (5/25 leak lesson — never bash -x token-minting scripts), run `~/bin/gh-fleet-token.sh` in a foreground session and capture stdout/stderr but redact any token-shaped strings before logging. 2) If it fails: check 1Password `Github (Coding Agent)` PAT entry validity, check App-ID 3379493 / Installation 123990571 token issuance via curl to api.github.com/app/installations/123990571/access_tokens with a generated JWT. 3) If it succeeds now (transient): note the duration of the outage and update LEARNINGS.md. 4) Open new lesson: fleet-task triggers are one-shot — file fresh task each occurrence, or build a recurring cron-driven check. Pick up next heartbeat with bandwidth.