meeb
commented
1 month ago I've been pondering some sort of soft metadata retry method where if a download fails enough to be marked as permanently failed on the first permanent failure it'll clear then re-download its metadata. This should be about a day later which would likely account for issues like the one you've described. The only downside here, other than even more complex retry logic and some additional database counters, is that if you have a mass failure it'll schedule downloading a significant amount of metadata again which could cause issues for some people.
This has been on the cards for a while. I've also put this on the back-burner pending the hopeful replacement of the worker system given this feature would need to be redeveloped after replacing the worker system.
After adding the ability to download live stream VODs the problem arose that many of the long streams aren't yet finished encoding by YouTube into the desired format and are skipped because of "no machting formats". It would be great if those videos would be retried to get download with the same logic as if YT-DLP runs into an error while downloading (maybe up to 5-10x times).