By chance, has anyone managed to resolve this type of problem (when the cluster is shut down suddenly)?