It is something to do with fixing a problem, and for that it wants to write a new database, but the old database is locked. I think I know now what causes this problem, but I do not have an easy fix.
The backup is keeping a connection to the database open, but the repair process wants to replace the database with another version. When attempting to move the database file while a connection is open, you get this error.
The fix requires a change to how the backup process is using the database, which would allow this in-place update. I am working on that (among other things) in the concurrent_processing branch.