Observed on kubernetes many times, @onursatici and I suspect the issue is linked to the k8s watcher disconnecting due to a k8s api server flake.