This also entails a "kill job" response to the workers' status updates, and a way of handling it in the JobFetcher code