Results of more testing:
In several steps I deleted versions back beyond the last good run of the job. While processing, the GUI displayed a lot of files that are ages old, and I wondered why they should be considered at all.
While the DELETE jobs seem to have run OK (except for the glitch with the version names), all backup jobs failed with a closed connection, as before. The progress indicator in the last run even displayed a ridiculous state before the job timed out:
While it usually hung at some 380 files and 0 bytes to do, this time it even went far below zero …
I’m glad this isn’t my only backup path for my data. Guess there is no way but to ditch all this crap and create a new job from scratch.
Let me repeat that 2 other jobs reading from the same NAS and writing to the very same destination cloud space continue working like a charm.