Since our activities are "statically" defined (A -> B, forever) it's possible (and probable) that some simulation activities may terminally fail. We should think about alerting / backoff / other ways in which we should handle this scenario.