Possible reasons:
a. New objects are still being replicated to the destination cluster. This could just be because:
* a large number of objects have recently been written
* the communication pipe is too small for the amount of data being written
* the replication task is paused (check admin)
* replication task is set to run in a time window (check admin tool)
b. Small differences can come from the MatrixStore system files - small files that are used for admin purposes and which aren’t replicated
c. Someone has directly deleted a number of objects at the destination cluster
d. Someone has directly moved some objects at the destination cluster after they were replicated
e. The source cluster could have a number of deleted files in its trash can