Currently, we add a 10 second sleep between bucket rebalancing to allow clients to detect vBucket map changed. In cases where there are larger number of buckets, it can add up (i.e. 30 buckets X 10 secs = 5 min).
At the very least, we can reduce it to a lower number, as 10 seconds might be too large. But ultimately, sending an unsolicited notification to all clients can be a more robust solution.