You are correct about the -0.5dB in the VAD. It starts the process of making sure there are no samples that exceed 0dBFS. Keep in mind the audio is captured in fixed point (PCM) format then converted to floating point just before processing. Then, after processing is complete, the floating point samples are converted back into PCM samples. This conversion back to PCM is where clipping may occur because floating point samples may exceed 0dB but fixed point samples may not.
We are currently investigating ways to take advantage of Windows floating point audio capabilities. However, most audio files are not floating point and most output devices cannot take a floating point signal.
The new headphone enhancer is a nice little feature for providing lightweight (for the CPU) stereo field control and customization for headphones. It provides a few easy presets for mixing left and right channels for more comfortable listening (especially for those old records where they put the drumset the left and the voice on the right) and lifting the stereo image. This effect is applied along with your favorite headphone profile.
It is currently available to SDK customers and we will be putting it into DPS 2.0 shortly.