Description
Currently we get the dcp docs estimate at the start of a rebalance by making a STAT call for every vbucket involved in the rebalance. In an marginally underprovisioned system, this can take a substantial amount of time, blocking rebalance progress, among other things, as seen in MB-60118.
We should investigate if we can improve the performance of this, for instance with pipelining or making one request for all the vbuckets at once.