That's cool. Thanks! I'd still argue that you want to provision N+2 capacity, so you can withstand a planned event and an unplanned one at the same time, witho

That's cool. Thanks! I'd still argue that you want to provision N+2 capacity, so you can withstand a planned event and an unplanned one at the same time, without having to go through manual tweaking. I'd leave that for truly exceptional cases, such as when the whole cluster is hit by some nasty bug or has entered a spiral that calls for drastic measures.