Hi Alex. My understanding could be wrong, but this is the scenario that I'm
trying to avoid:
1) Machine 1 has the original data and backs it up locally to /backup
2) Machine 2 rsyncs /backup to local /backup every day
3) Machine 2 is rsyncing /backup, and right then machine 1 dies for good (fire,
flood, apocalypse)
At that point in time, machine 1 is dead and /backup in machine 2 is not a
valid rdiff-backup directory, so you can't restore your data from it anymore.
If Machine 2 was instead using rdiff-backup on the original sources, /backup
would have an incomplete backup, but previous data would not be lost, so you
can restore from it even if it was interrupted.