e.g. farming out to workers which are not part of the cluster, or using a cluster with multiple roles for managing parts where scaling out happens using remoting and clusterClient again, plus whatever we can find in the wild