Shane, can you check the fix: https://www.dropbox.com/s/k91cqf6atmjbukh/HangfireFixedRace.zip?dl=0 ? I’m not able to reproduce the source bug neither in office, nor at home 
The problem was related to a very quick worker that picked up a background job that was not fully initialized for the enqueued state. I’ve fixed it by waiting for a full initialization (with timeout) before trying to change the state to Processing. If timeout expires (or when process is killed), the job simply moves back to the queue.