When starting a reconciliation in KafkaAssemblyOperator, we first collect the cluster description based on the existing Pods, StatefulSets, StrimziPodSets etc. right now, while supporting both StatefulSets and StrimziPodSets in parallel, we create the description based on the UseStrimziPodSets feature gate. If the feature gate is enabled, the description is collected from the StrimziPodSets. If it is not enabled, it is created from StatefulSets.

This works fine in most cases, when the feature gate is stable. But there might be some edge cases where this might cause problems while enabling or disabling the UseStrimziPodSets Feature Gate. The current implementation means that in the first reconciliation after the gate is enabled or disabled, the reconciliation will start with empty description because the new StatefulSet / StrimziPodSet does not exist yet. This could cause issues for example if storage is changed at the same time the feature gate is changed.