A really cool feature for us would be if you could run long running jobs in a cluster (multi server) setup. One example here is Azure webjobs where you can interop with servicebus queues. If a long running job becomes unresponsive then kill it and reschedule it. This would offer awesome flexibility and a way of distributing load if you a lot of servers.
You would only have to implement the actual logic and then we can trivially extend it with extra attributes supporting various workflows as long as the job is guaranteed to always be running on any given server.
/cc @odinserj