[CI] Premerge buildbots not restarting automatically

[boomanaiden154]

Since the buildbot workers run inside k8s pods, they get killed when a maintenance window starts. They don't necessarily restart correctly and need to be manually restarted again to reconnect to the buildbot master.

This should be fixed so that it is automated and no intervention is needed from the oncall.

[RoxaneFis]

workers often don’t reconnect cleanly after pod evictions, so adding readiness/liveness probes and a PodDisruptionBudget can help automate recovery. For a deeper fix, our AI SRE at Anyshift analyzes k8s events and logs to explain failures and handle recovery without oncall effort