We should try to avoid unnecessary patching of resources being deleted (but their deletion being for example blocked by finalizers for several minutes). One of the examples where this might cause a problem is when you delete a PVC -> it should not be deleted while being used by a pod. But because the operator patches the resource, it removes the finalizer and gets it deleted while still in use. In some environment, that seems to also wipe the PV and the data volume and fail the pod. In others, it seems the PV deletion will not happen until the pod rolls. But none of the cases is desired.