The instance was struggling a lot when I woke up this morning. For some reason we had more than 20k jobs in queue. 10h later, there are still 8k waiting to be processed. I suppose the fact that Wasabi is shit doesn't help.
On the same note, I'm aware that media attachments don't work most of the time. I'm trying to find a solution, which involves switching object storage providers, and we're talking about hundreds of gigs so it involves time and money. Sorry for the inconvenience.