restarting a node with more than 2M items could result in loss of 0.001 percentage of active or replica items due to timing issues

4 node cluster, 3 million items

• reboot one node, wait for warmup
• eject any node
• add back
  Noticed item count dropped and active/replica numbers mismatch.

Chiyoung suspects this is due to checkpointing bug fixed in 2.0 latest and plans to back port fix so that we can test with new build.