The basic problem with the poly voice manager is still the same. The whole poly system is very hidden away inside SM's engine, so there no way to change the voice allocation unless the developers re-write the MIDI-to-Poly primitive.
Earlier in the thread, it was mentioned that making a custom voice manager using "Green" and mono streams might be possible.
I have done quite a bit of work on that method, but it is quite cumbersome - 4 or 8 voices is manageable, but it gets very messy the more voices you add, and CPU load is higher because the voices are always playing, even when you don't hear them.
However, I think it can be useful as you can also get around other "preset" behaviours of the stock manager - for example, normal poly voices always reset oscillator phase at note-on, whereas a "blue/green" system allows for genuinely free-running oscillators. So, if there is a very particular behaviour that you need, and don't mind the CPU and voice-count compromises, it might be worth trying.
You can see some of my experiments with a four voice version on the Green Voice Manager thread
- and I'm currently working on an 8 voice synth using very similar modules.