Worker doesn't shut down on failure

[gentlementlegen]

The workflow keeps running indefinitely on error, which eventually times out due to GitHub limits. I believe we can have retries but eventually should stop after an amount of failures.

See https://github.com/ubiquity-os-marketplace/ubiquity-os-kernel-telegram/actions/runs/11400116637

[Keyrxng]

@gentlementlegen I tried a little earlier to repro the same error auth_key_duplicate but couldn't. By any chance do you remember what you done to cause it? Ok if not I plan to force errors but just to streamline things I guess.

[gentlementlegen]

@Keyrxng I think this happens when multiple runs start at the same time. The cause is that the worker have different IPs but use the same key.