Suggestion: 2 or 3 dimensional virtual voice locations

You should implement a feature for headphone users (and speakers depending on the surround sound setup) that allows the user to choose a virtual direction to hear another user in the chat from. e.g. you could set one person to sound like they were coming from directly forward, another could sound like it's coming from 45 degrees to the right, maybe up and down, etc. 

This would be very convenient to help identify who is talking, and separate voices when multiple people are talking at once. In the headphones, the voice orientation would obviously rotate with the user's head because headphones generally don't have any sort of accelerometer or inertial tracking, but you could implement that as well for any headset that does have this (if any), or even some sort of clip on accelerometer. This part is less important than the separation of voices, but is something to consider.