We should document the edge case that exists with Escalator in that when there is a consistent amount of base load and Escalator always maintains a certain amount of nodes, it will never cycle out the oldest nodes first, as they'll always be used to run workloads.