On memcached kill, bucket on master node doesn't warms up until the swap rebalance of other 2 nodes are completed. Rebalance doesn't abort while it should be aborted and completes with bad_replicas

1. Create a 4 node cluster
2. Create required buckets and collections.
3. Create 5000000 items sequentially
4. Create 5000000 items sequentially
5. Update 5000000 RandonKey keys to create 50 percent fragmentation
6. Create 5000000 items sequentially
7. Update 5000000 RandonKey keys to create 50 percent fragmentation
8. Start doc loading in async manner
9. Rebalance in with Loading of docs. Abort and resume rebalance at 20, 40, 60, 80%
10. Crash Magma/memc with Loading of docs
11. Rebalance Out with Loading of docs. Abort and resume rebalance at 20, 40, 60, 80%
12. Crash Magma/memc with Loading of docs
13. Rebalance In_Out with Loading of docs. Abort and resume rebalance at 20, 40, 60, 80%
14. Crash Magma/memc with Loading of docs
15. Swap with Loading of docs. Abort and resume rebalance at 20% by killing memcached on master node. But rebalance doesn't get aborted and bucket on master node never comes out of warmup until the ongoing rebalance is finished.

Note: In any of the above rebalances master node is not involved in rebalance operation. It is just being used to abort rebalance by killing memcached on it.