@drwtsn32 The USN journal only records file and folder names, not full paths. But each name is also associated with a parent entry, which is the containing folder name. So this code resolves this chain all the way up to the root, building up the full path name.
But since the USN has only limited size, parent entries get ejected, and it becomes impossible to generate the full path. This happens if backup jobs happen too infrequently, in which case the USN journal size doesn’t contain the entire history data since the last time it was checked.
Except for increasing journal size, I don’t see how this can be fixed.
But: instead of reverting to a full scan, Duplicati could just perform the full scan for the affected parts, and still USN where it was able to obtain the change information. Actually, I don’t remember if I already did that, but I don’t believe so. In this implementation, the slightest USN failure triggers a full scan.