Data loss seen with graceful failover on 4.6.4

Steps to Repro
1) Created a 4 node cluster of 4.6.4
2) started 3M data population using cbc-pillowfight
3) When 2 in progress started graceful failover on 4.6.4 node(172.23.98.196)
Data loss was noticed after step 3.
4) Installed 5.1.1-5716(172.23.98.196) on the node and did full recovery and rebalanced.

One doc was missing. Key - 00000000000000948466. It belongs to vbucket 910.

[root@CentOS7 ~]#  cbc hash -U  couchbase://172.23.98.195/before_upgrade_bucket -u Administrator -P password 00000000000000948466

00000000000000948466: [vBucket=910, Index=3] Server: 172.23.98.196:11210, CouchAPI: http://172.23.98.196:8092/before_upgrade_bucket

Replica #0: Index=1, Host=172.23.123.251:11210

[root@CentOS7 ~]#  cbc-cat -U  couchbase://172.23.98.195/before_upgrade_bucket -u Administrator -P password 00000000000000948466

00000000000000948466 The key does not exist on the server (0xd)

[root@CentOS7 ~]# 

However tried repeating the test with other 3 nodes that were remaining in the cluster, I was not able to reproduce the issue and did not notice any data loss.

Live server available for debugging - http://172.23.98.196:8091/ui/index.html#!/logs

Logs attached.