As mentioned in the first post by the OP, synced mapdb persistence file happened to get corrupted during his test. Corruption of synced files never occured during all my tests but I did not push the workload very much because I was focused on replicating the issue of RawCopy always needing to perform e2fsck before set-partuuid.
In principle, by increasing the workload using my test scripts the rsync can be tested in a more demanding context. Possibly mapdb is more prone to corruption because many items are related to a single file.
