Manually copying new data to existing replica

In our environment the DPM servers back up file servers over the WAN.  We do an initial seed via USB to prevent massive amounts of data from having to transfer over the WAN.  This works great.

My question is regarding adding more data manually to the replica.  We sometimes have situations where a large amount of data needs to be placed on the file server (say 1TB).  This is too much to sync over the WAN.  So what we have been doing is something like this:

1. Mark replica as inconsistent
2. Manually copy the data to the DPM replica from a USB copy (similar to initial seed)
3. At the same time manually copy the data to the file server
4. When both copies are complete, run a consistency check

This usually works fine.  But sometimes the act of copying new data to the DPM server replica volume exhausts the shadow copy volume. Shadow copy free space is reduced at the same rate that new data is copied into the replica volume.

I want to understand why that sometimes happens and other times does not, so that we can adjust our practice so that it never happens.  In all cases we are copying NEW data, usually to a new folder.  We are not overwriting or updating existing files (if so I would understand the increase in shadow copy volume usage).

Any insights would be appreciated.