I've noticed on smoketest that after long periods of time (days) the stream-loader and url-fetcher can sometimes start to fail to pick up new jobs from Redis--but are still running. I think that this is happening because the event loop inside of each type of job needs to log and catch when an exception occurs. I think we need to look specifically at UrlFetcher, VideoFetcher, StreamLoader and SearchLoader.