Pages deploy can silently stall for days behind a run stuck in 'waiting' #203

Scheduled Freenet Deploy run 37512153827 (2026-10-06, e3556da) left its deploy job in waiting on the github-pages environment for six days. The environment has no required reviewers and no wait timer (only a branch policy), and pending_deployments listed no reviewers, so nothing could ever approve it.

Because deploy uses concurrency: group: pages, every later deploy queued behind it and was cancelled when the next one arrived ("Canceling since a higher priority waiting request for pages exists"). So freenet.org didn't update from 2026-10-06 until 2026-10-12, which covered #201 and #202, and nothing failed loudly. timeout-minutes: 15 doesn't help, because a job waiting on an environment hasn't started.

Cancelling the stuck run on 2026-10-12 unblocked the queue, and the next deploy (38291798728, 137e439) succeeded.

The root cause looks like a GitHub-side glitch, since no rule should have held the job. Possible guards:

• A scheduled check that fails (or notifies) when the latest successful github-pages deployment is older than ~2 days, or when any Freenet Deploy run has been waiting for more than an hour.
• Or have the daily run cancel any deploy run of this workflow that has been waiting for more than N hours before it queues its own.

[AI-assisted - Claude]