Possibly related to HIVE-1545, but assume we discover we're approaching the limits of Hive in a kcp server. How could we rebalance some of those ClusterDeployments to another kcp server and another set of Hive controllers?
Explore migrating keys in etcd. (does this bypass the risk of API deletions? does it break caches?)
Could this kind of rebalancing be built into kcp?