Partition handling currently relies on every cache being defined and running on the coordinator. We could relax this, as the partition handling configuration can be easily added to the cache join command.