How well does scaling work on the GraphCool pay as you go shared cluster? The nature of our service has very little traffic for a few weeks then pretty large usage for 30hrs or so. Will it scale up and down automatically?