Lowering priority: while I think there will be lots of research interest in different types of padding machines, I don't think the actual machine updates will require update via consensus until we start producing many different types of machines for different quantities of spare consensus overhead, which is a long way off.
I think #29203 (moved) plus #29098 (moved) plus the existing global consensus padding rate limits are way more immediately useful for reacting to different amounts of network load than swapping machines in and out entirely on the fly, at least until we learn more.