Data loss is observed due to magma rollback

1. 4 nodes cluster: Node1(0-255), Node1(256-511), Node1(512-767), Node4(768-1023)
2. Load 100k in the bucket. Make sure all of them are persisted.
3. Turned off persistence on node1
4. Load 100k on Node1.
5. Kill memcached on node1. All the data loaded in Node1 in step 4 vanishes
6. Step 4 will trigger rollback on all other nodes.
7. Data should be 100k in the bucket but there are few items missing. Items present=99,973

Logs:
https://cb-jira.s3.us-east-2.amazonaws.com/logs/magma_rollback/collectinfo-2020-01-03T063431-ns_1%40172.23.105.220.zip
https://cb-jira.s3.us-east-2.amazonaws.com/logs/magma_rollback/collectinfo-2020-01-03T063431-ns_1%40172.23.105.221.zip
https://cb-jira.s3.us-east-2.amazonaws.com/logs/magma_rollback/collectinfo-2020-01-03T063431-ns_1%40172.23.105.223.zip
https://cb-jira.s3.us-east-2.amazonaws.com/logs/magma_rollback/collectinfo-2020-01-03T063431-ns_1%40172.23.105.225.zip