I have just found the following bug / regression:
if a file at destination is already existing, but was modified at source.
If a md5 match is found with another files at destination.
The file that is tried to be "remote copied" with just be the first letter of the "other file".
So the remote copy will fail, and there is a possibility of a "wrong copied" file if there was files with a single letter at destination and that it did match the first letter of "other file".