I recently experienced a situation where, after upgrading clickhouse-operator from 0.25.3 to 0.25.6, the operator ended up deleting both StatefulSet resources from my cluster (1 shard 2 replicas) while reconciling changes. Another similar situation occurred where the operator deleted one of the StatefulSet resources and managed to reconcile/update the other.

You can find the logs from clickhouse-operator in this gist

I started looking into what could have caused the issue with Claude's help, but i don't fully trust the conclusions. Claude also wrote up a "bug report", which I have cleaned up a little and pasted into this gist. If you prefer, I can update this ticket and inline the contents of claude's "bug report" here.