@p @amerika @FranciscoScaramanga @GummyNerds @PNS @eriner @sun You can't do anything about it on the Mastodon side, I think, without extensively modifying how the queues work.

On the Pleroma side, I talked with feld about some kind of queuing for those delete jobs and just sending "success" back without immediately deleting the object. The idea was that it would periodically execute those jobs with a much lower priority until the queue is flushed completely and not hammer the DB like it does right now. But that idea sort of faded away after some time as he worked on more important things. Can't blame him.

Of course Akkoma, in typical Akkoma fashion, can't even cope with the load, as the MRF pipeline was modified to exclude Delete activities.