the voices are mono, by default L+R inputs are summed for each voice but you can easily change the per-voice input mix in the script.
for 2.0 i am experimenting with a couple different methods for syncing voices and handling stereo files in the engine. (UI options in the MLR script are another matter.) also i think there will be 2 large buffers instead of 1.
which is to say, one way or another you should be able to use pairs of voices as stereo pairs, for live recording and/or buffer playback.