Description
It's been observed many times (at customers) that the immediate cleanup of dead vbuckets on all nodes of a cluster can cause performance issues...
These issues are made even worse when the cleanup happens after a rebalance failure because not only will replication kick in and stress the cluster further, but the most liekly times for a rebalance to fail are when the cluster is already under stress and we don't want to put even further load on it.