My workaround for this was to put in a temporary schedule to start the backup at 12:05AM. If it has 403’d, it starts up again, if it hasn’t, then the backup continues running.
This does waste time with 
figuring out what it has already uploaded, but even with ~10TB presently backed up (and ~6TB to go), I still only get ~2 hours/day of uploading, which means it can spend a lot of time waiting on metadata.
