#3390 @swamp/aws/ec2 has not published since 2026.09.29.1: its 23.4MB archive's publish job is killed after ~14 minutes

Found while publishing #3325. The registry's current @swamp/aws/ec2 is 2026.09.29.1. Main's 2026.10.11.1 never published, and neither did #3325's 2026.10.13.1, including when dispatched alone (publish.yml run 2056, task 42160: started 19:11:10, failed 19:25:14 UTC).

What we know

• swamp extension push --dry-run passes locally in about 39 seconds, warnings only. The archive is 23.4MB (about 120 models).
• Publish jobs on this repo also fail on a five-minute grid (18:25, 18:30, 18:35, 19:25 UTC) after 12 to 23 minutes, which looks like the server reaping tasks whose runner stopped reporting. Today 11 of 28 chunks of run 2052 failed that way, and one of five in run 2054; all of them succeeded on a re-dispatch except ec2.
• So either the ec2 push takes longer than the jobs survive (a slow upload or registry-side processing of a large archive), or it hangs.

Next

• Get the job log for task 42160 (the logs endpoint was not reachable with the fj token) or push ec2 by hand with --verbose and time the upload and confirm steps.
• If it is size: find the registry's limit or timeout; consider splitting @swamp/aws/ec2, or making the push resumable.
• Separately, find out why publish tasks are being reaped (runner heartbeat, job timeout) — it costs a manual re-dispatch on every large publish.