part of me wants to give him the benefit of the doubt; he's talking about latency and I know from experience there's an undesirable "clicking" artifact produced when the sample buffer is empty (ie, the application isn't producing samples fast enough); this is caused by the frequency abruptly dropping to 0Hz. It could arguably be described as a "hard edge". So maybe he's in a situation where libc's memcpy was a bottleneck and was preventing new samples from being sent to the audio hardware fast enough.

But the part where he says that the c++ new[] operator sounds better than malloc is just ridiculous. I'm not good enough at mental gymnastics to make that one seem remotely plausible.