I'm not sure that has been discussed already. I would like to choose how voice are distributed in split mode. I've had some synths for which you could have 2/4 and 4/2. As I have the Rev 2 8voice (will upgrade to 16 very soon) I would like to have 1 or 2 for the left and the rest for the right hand side.
I think 2/6, 4/4, 6/2 for 8 voice and 2/14, 4/12, 6/10, 8/8, 10/6, 12/4 and (you guessed it right) 14/2 for the 16 voice, is not that much to ask for.
If I'm right, this is not possible because of the hardware... the more I look at the REV2, it seems like you should see the synth as two independant 8-voice synths... one for each layer... when being in 16voice mode (only using layer A), I bet that layer A is automatically copied to Layer B so their sounds are exactly the same, and then an internal kind of "Polychain" method is used so that it seems like you have one synth with 16 voices.
If this is correct, then the two layers are hardwired each to their own with their own FX slot... thus it would be impossible to use any voices from one layer, on the other layer, as they are hardwired to the given layer.
If I'm not correct on this, DSI may correct me... but I'm pretty certain this is what is going on... and if I'm not mistaking, I bet this is the same for both Prophet 12 and Prophet X as well.
The reason I believe that the sound on Layer A is normally copied to Layer B on a preset using only layer A is, that the edit buffer dump bug that I've found earlier, seems to forget to copy the layer from A to B when it receives a dump that should be in 16voice mode... when this happen, the 16 voices that will cycle have the first 8 keystrokes being the right sound, but the next 8 in the cycle is whatever is stored on layer B of the dump... so if you say; dump a preset that use only layer A, and layer B just contain the "basic Preset B" sound, then the first 8 keypresses gives the intended sound, but the next 8 gives you the "Basic Preset B" sound... and so on and on and on...