Rollout operator cannot recover a StatefulSet from CrashLoopBackOff

[grobinson-grafana]

Problem

I've observed an unfortunate behavior with the rollout operator where it cannot recover from a failed rollout. The sequence of events works as follows:

1. The StatefulSet is updated with a new image or a new configuration of some kind, and the rollout operator starts to roll out the StatefulSet.
2. The first pod(s) do not become healthy / enter CrashLoopBackOff.
3. Rollout operator pauses the rollout because it is waiting for pods to become healthy.
4. A human notices the issue and fixes it by updating the StatefulSet.
5. Rollout operator gets stuck and does not apply the fixed StatefulSet to the crashing pods.
6. Crashing pods remain in CrashLoopBackOff until a human manually deletes them, and only then does the rollout operator apply the new StatefulSet.

Solutions

Can the rollout operator work like the RollingUpdate strategy where it restarts the rollout from the beginning even if the first pods from the previous rollout are unhealthy / crashing?