I was a bit suspicious of your assertion that no corruption is possible if it fails during the merge, because I noticed that the full back itself was modified during the merge and I thought "well, that's bound to go wrong!" leaving me with a corrupt full backup. But I've tested it by disconnecting the machine being backed up during the merge, actually during the phase when the full backup is being modified (growing in size), and whilst I got a scary message saying "process finished with errors", when I reconnected and started another backup I got the reassuring message "completing previously interrupted transform", followed by "merging oldest incremental backup into full backup file", followed by "merge (%)", and then the merge completed correctly. So, I'm very impressed! Thank you for such a great invention.