Yeah I don’t buy that. It’s a long pipeline and there’s almost definitely more than 3 milliseconds of jitter. I’d wager that success/failure sounds are quantized to audio buffers that are a good bit longer than 3ms.
This is a deep and interesting enough topic that you could publish a pretty good and valuable paper on what the JND is for poor, average, elite, and “professional” players is.
If the line is 3ms for pro I will gladly bet the over!