Hm, I think you’re right, these cases aren’t explained in the manual. In addition, we need to rethink the error messages check prints. For example, the message “file <1234> is not referenced in any index” is not critical at all, it could easily be suppressed and the next run of prune will take care of it.
Just re-run the backup, it’ll pick up approximately where it left of. A small amount of duplicate data is left in the repo from the previous run, most already uploaded data is reused automatically. The next rune of check will likely print messages like the one above, the next run of prune will take care of the additional files.
Restart the restore process from the beginning in a new directory (or remove the files). There’s currently no resumption for restoring files. We may eventually add this, but it’ll require some complexity, and, to be honest, it’s down below somewhere on our todo list.
I hope this helps! I’d be happy to merge this in a PR, adding some text to the manual and/or improve the message check prints.