Rebalance hang in large cluster test with "mem_used" stat underflow issue

Install couchbase server 1.8.1-847 on 12 node in ec2 with 7.5 GB RAM and 100 GB disk
Create a 10 nodes cluster and load 40 million data into default bucket with size from 256 to 512 bytes
Maintain load around 7 K ops
Do rebalance in and out in large cluster test
In one rebalance test, rebalance hang near complete with error:
Hard Out Of Memory Error. Bucket "default"

Diags of cluster:

https://s3.amazonaws.com/packages.couchbase/sandbox/ec2-23-20-53-38.compute-1.amazonaws.com-8091-diag.txt.gz
https://s3.amazonaws.com/packages.couchbase/sandbox/ec2-23-22-45-111.compute-1.amazonaws.com-8091-diag.txt.gz
https://s3.amazonaws.com/packages.couchbase/sandbox/ec2-50-16-17-17.compute-1.amazonaws.com-8091-diag.txt.gz
https://s3.amazonaws.com/packages.couchbase/sandbox/ec2-67-202-49-238.compute-1.amazonaws.com-8091-diag.txt.gz
https://s3.amazonaws.com/packages.couchbase/sandbox/ec2-72-44-47-164.compute-1.amazonaws.com-8091-diag.txt.gz
https://s3.amazonaws.com/packages.couchbase/sandbox/ec2-107-21-88-105.compute-1.amazonaws.com-8091-diag.txt.gz
https://s3.amazonaws.com/packages.couchbase/sandbox/ec2-174-129-130-202.compute-1.amazonaws.com-8091-diag.txt.gz
https://s3.amazonaws.com/packages.couchbase/sandbox/ec2-184-72-67-54.compute-1.amazonaws.com-8091-diag.txt.gz
https://s3.amazonaws.com/packages.couchbase/sandbox/ec2-184-73-111-82.compute-1.amazonaws.com-8091-diag.txt.gz
https://s3.amazonaws.com/packages.couchbase/sandbox/ec2-184-73-112-165.compute-1.amazonaws.com-8091-diag.txt.gz
https://s3.amazonaws.com/packages.couchbase/sandbox/ec2-23-21-17-242.compute-1.amazonaws.com-8091-diag.txt.gz