Since the last cluster update, we’ve been running our staging instance on a separate node pool which uses AWS Spot instances. It causes the staging instance to be quite unreliable. We’re testing it in order to find balance between price and usefulness.