This was mostly properly scaling down things that didn’t need to be up all the time (as far as I understand it, I wasn’t part of the effort). We do already have a system that shuts down services when they’re not needed (based on tags, so presumably sort of the same thing).