Embed Notice

@puniko@mk.absturztau.be if you want to experiment more, you can try setting deliverJobConcurrency to 512 (on top of deliverJobPerSec being -1)
do be a bit careful though since it's IO-sensitive (so you might need to lower it a bit based on your workers, you can easily tell based on whether or not you get a bunch of timeouts in logs after a delete)