The parent tasks's T428867#12013155 uncovered a backy2 bug where a differential backup fails if it can't find enough matching pre-existing data in the last snapshot from which calculating the diff.
Instead of falling back to a full backup an exception is raised.
In this case wmcs-backup retries 3 times, but this is a deterministic failure, so it failed all 3 times and then made the whole backup run fail.

This should ideally be fixed in backy2 (that is mostly unmaintained) and also in the way wmcs-backup handles the failure. Ideally it should continue with the other volumes and fail the whole unit only if the % of failures is over a certain threshold. It could still exit with a failed status (to trigger monitoring once T428873 is fixed) for a single failed backup, but it should continue to backup all the others.