As Lindenberg Software Backup only transmits deltas from the client to the server, it is of utmost importance that client and server agree on what is the current state on the server. Basically it works as follows:

When the client starts a backup, it checks with the server whether they are synchronized. This is done by checking the modification time of the image and the hash files on server and client - if all three files show the same modification time, client and server are in sync, and the client just sends deltas. If not there are two possibilities: if image and hash file on server are not consistent, then the server automatically rebuilds the hash file locally and denies any client backup until that is resolved. In fact the server could do this on the fly, however the process is essentially IO bound and would significantly slow down any backup. If hash file and image on server are consistent, then the client checks with the server which parts of the hash file changed and uses any modified server hashes as input for the change detection.

In other words: if a backup is interrupted because e.g. the client crashes, runs out of battery, or the network is interrupted, this is usually resolved very quickly and causes only a slight delay during the next backup. If the server detects an inconsistency, then it will resolve that automatically during service startup or as it detects it, and may decline backup service to the client until that is resolved.

Unfortunately the backup service might not detect an inconsistency in case of a software bug or a server crash. If you suspect a bug in backup service, please report it (see also Bug Bounty Program). In case you hit a server crash due to power failure, blue screens, etc., or by any other cause, the easiest way to enforce a consistency check is by deleting the hash files on the server and restarting the backup service.