There is an issue related to running Resque on Heroku and Resque's ability to clean it's state information for workers. This is mostly a cosmetic issue however we got to a point where the Resque Dashboard would timeout because it was trying to display hundreds of workers it thought were sitll running.
This is a heavy handed approach to cleaning up. We are finding any worker that was stared longer than a day ago and manually calling the unregister
method. Be advised this will throw DirtyExit errors.