The frontend operator has a frequent, but not constant, condition where all pods will be created and terminated in a storm. We assume there is a reconciliation issue with the way we create the config hash that makes it non deterministic. When any new change is applied to a frontend, we see this pod storm. Pete may have a fix so we shall investigate. 

To reproduce: Run > 1 Frontend resource in ephemeral or local. Make a change to any of the frontend resources. Watch the storm, probably.