And that pitches out the window fault tolerance or replicability unless nontrivial developer work is dumped into it. If the disk goes farming in your Kimsufi instance, you are looking at a bad day.
Sure, if the machine dies, my services die. But it's still dirt cheap.