The computer’s internal audio is 44.1, the audio CD standard. That’s the problem. The processor has to resample the audio on the fly to play back the sound through the speakers, while it’s capturing at what’s most likely a different sampling rate. This is a bad idea. In older systems it was buried in capture presets because slower processors would actually cause sync drift during capture. That’s what Ralph Fairweather found when he was QA for the original Final Cut Pro. Modern systems with faster processor speeds probably don’t have this problem, but I could well see it causing odd sound problems though. I don’t see the point of risking it myself.