I was informed about something going wrong here and investigation showed that it's worse than I thought. My Contabo Object Storage started claiming it's full, even though it's only half-full. I opened a ticket to tell them to fix their shit, but knowing their speed, it might be a few more days :-/
That's what I get for choosing the budget services...
Maybe as a stopgap, you can try cleaning up the auto-generated variants (alternate-size thumbnails) that pict-rs has stored? Might get you under whatever threshold was tripped while supports works out the main issue.
DELETE /internal/variants : Queue a cleanup for generated variants of uploaded images.`
If any of the cleaned variants are fetched again, they will be re-generated.
My instance is fairly conservative with media uploads and cached thumbnails, but it shaved off a few gigs that had accumulated over 6-7 months. Heads up that the call to that endpoint will return immediately as it just queues the job in pictrs.
Unfortunately I already tried deleting some backups I had in there. After 10Gb freed, it still gave me the same issue. So I think it's some bug on their end.