[Query-XDCR] Duplicate document returned from query with stale=false causing additional rows in the query

http://qa.hq.northscale.net/job/centos_x64--107_01--rebalanceXDCR-P1/57/consoleFull

[Test Error]

======================================================================
ERROR: swap_rebalance_replication_with_view_queries_and_ops (xdcr.rebalanceXDCR.Rebalance)
----------------------------------------------------------------------
Traceback (most recent call last):
File "pytests/xdcr/rebalanceXDCR.py", line 283, in swap_rebalance_replication_with_view_queries_and_ops
[task.result(self._poll_timeout) for task in tasks]
File "lib/tasks/future.py", line 160, in result
return self.__get_result()
File "lib/tasks/future.py", line 112, in __get_result
raise self._exception
QueryViewException: Error occured querying view default0: expected number of rows: '70000' is greater than expected 70001

----------------------------------------------------------------------
Ran 1 test in 2215.179s

[Test Steps]
Source: 10.5.2.228, 10.5.2.229, 10.5.2.230
Destination: 10.5.2.231, 10.5.2.232, 10.5.2.233

1. Setup 3-3 node Source and Destination Cluster.
2. Bucket: default, mode: CAPI mode unidirectional xdcr.
3. Load 1M items on Source.
4. Create DesignDoc: ddoc1 on Source and Destination
5. Created views: default0, default1, default2, default3, default4 on Source and Destination.
6. RUN Parallel
i. 30K and Delete 30K items on Source
ii. SWAP Rebalance one node on Source (Swaped 10.5.2.230 with 10.6.2.192 on Source)
iii. Running view-queries on all views on source and destination until rebalance is finished and mutations are finished with stale=false e.g.http://10.5.2.231:8092/default/_design/ddoc1/_view/default1?full_set=true&stale=false)

7. Wait for re-balance to finished.
8. Again run queries on all views with stale=false to verify expected rows=70000.

Error Expected rows: 70000 and Actual: 70001