I was wondering whether the execution of a fire and forget job can be guaranteed to execute only once (off course there can be retries, but only 1 successful). I was thinking of maybe Hangfire and the job success status updating be part of the same transaction as the work that is done in the database. If this is possible, i can save myself the work of making my jobs idempotent 
I do not expect this to be possible, but just checking.
Also great project!