Mismatch on revision ids for deleted items with XDCR replication

Setup:
---------
1. Setup unidirectional replication between 2 two node clusters
2. Load 100k items on source cluster
3. 100k items are replicated to destination cluster( works as expected)
4. Delete 50k items on cluster1, 50k items are deleted on destination cluster as well.
5. Load 100k items using same prefix on source cluster.
6. Verify curr_items on source and destinaton - Counts as expected
5. Validate data : valid-keys and deleted keys via kv-store - Looks okay.

Error
----------
Verify revision ids for the deleted items.

• Establish a memcached client connection to source and to destination.
• Get the revision id information from both source and destination cluster for each deleted key.
• We expect the revision ids at both source and destination to be identical.

Seeing mismatch on revision ids on 200+ keys.

[2012-07-17 09:35:26,780] - [task:574] INFO - ******MISMATCH ON REVISION IDS **********
SOURCE-(3024256250212237, 0, 0)
DESTINATION-(3024256250212238, 0, 0)

• This job runs on Jenkins -
  xdcr-mod.unidirectional-xdcr.unidirectionalReplication.unidirectional_load_with_ops,replicas=1,items=100000,value_size=256,ctopology=chain,rdirection=unidirection,doc-ops=create-delete-create

Jenkins logs are not available, re-running this locally and will add the logs.

I would like to check with you on the cas/expiry flags/seqno alignment to verify whether we are checking the correct expected values.