At my job, the lack of high reliability/availability options for Redis was why we chose DelayedJob. This may have changed since we made that decision, but at the time there was no consensus on a solution in the Redis community.