Instead of forwarding the requester ref to the worker and replying directly after finishing work, a worker should report back to its master, who mediates the result back to the requester.

This way timed out work results could easily be discarded by the worker

a worker's doWork() should yield a Future[Unit] and then send itself the 'done' message in order to proceed

https://github.com/hensb/masterworker/blob/master/src/main/scala/de/uniwue/info2/masterworker/Master.scala#L42-L48

After a given timeout a worker should be flagged as 'timed out' and his work should be redispatched to someone else. If meanwhile the flagged worker returns his results, they should be discarded. A timed out worker should be removed from the master (maybe inform the worker, issue handshake-request,... ).