I think I read a thread like this almost 10 years ago when I happened to be doing some audio work, maybe it was the same one.

If memory serves, there was somebody saying that exiting the program cleanly sounds worse, you need to crash on a page fault reading the input buffer instead, and that's your end of file behavior.

Someone else said FLAC makes it sound worse, but I suspect they just couldn't figure out how to use libflac.

In other words they were bad programmers writing buggy, half baked shit, and they explained away their broken code as sounding better.

Even the snippet I see in the linked post, right off the bat they start their loop with WaitForSingleObject, that's kind of suspect; the first thing they do is enter the kernel to block at every loop. I suspect they aren't doing that in an optimal way.