Common reasons are:
a. Replication hasn’t completed for recent objects added. See any of the above.
b. Small differences can occur due to system objects.
c. The cluster holds partial objects for a period of time - ie - objects half written but abandoned. Sometimes there are a large number of partial objects if network comms are unreliable. The number of objects in this state can be seen in the admin tool.
d. Any of the reasons given above.