It is the blind claim of it being "plenty fast" that is resonated in this thread that I find problematic. The article describes how to implement queueing/app locks/pub sub, without providing any benchmarks on how the same implementations compare to redis under different loads.

If you are using redis and you don't care about perf you are probably doing it wrong. If you are using postgres and you care about perf you are probably doing it wrong. "Probably" because context is key.