OS X is very susceptible to bad clocking info from USB audio devices:
CoreAudio is a multi client system, each client can use a different buffer size (as long as it is smaller than the driver ring buffer) to the same audio stream, therefore each thread will need waking at different times. It is important to note that in OS X there is a disconnect between the actual driver data and the client user space data, you do not have direct access to the hardware driver as it was on OS 9. CoreAudio uses a push model where it pushes the data to the clients, same data, different sizes at different times, so it is predictive in it's timing and this is based on the driver data.
Timing issues are more general in USB interfaces and CoreAudio double checks the driver timings against an expected behaviour model, if the timings depart too much from the norm then CoreAudio will ignore them.
​What this means in the end is that the Client (Say A DAW) is woken at the wrong time to process the input audio leading to audio problems.
Windows ASIO does not work (usually) this way as the audio system is single client, unless you are running a multiple client ASIO driver for example the steinberg one. It would be interesting to see if the steinberg one suffers.