I know you can often get away with this kind of star grounding scheme at audio frequency, and often times it does work well, but I do get a bit squeamish when people really go nuts with it. While it's no tube amp, circuit impedances tend to be pretty high, and extra loop inductance and stray capacitance can sometimes lead to oscillation.
I'm not 100% clear on some of the details of your amp, but this is what I would suggest as a starting point:
For the twisted pair to be effective, the ground conductor in the twisted pair should not (inadvertently) shorted out by some other external ground connection. See my third image. I think I would avoid grounding each SIG_GND (L & R) + pre-amp GND to the star-point at the same time. Twisted-pair is great at rejecting magnetic field pickup, but if a parallel ground connection is made inadvertently, then your "tight twisted pair" becomes more of a "loose twisted pair", with a larger effective pickup loop area.
To me, seems the best solution would be to make one ground connection (to the star-point) somewhere on the pre-amp board. The SIG_GND connections on the power amp boards would be made through the twisted pair connection.
For shielded twisted pair, not much difference, but you can try attaching the shield at one end or both ends and see if you can perceive a difference. In the attached second image, I just tied on one end, to the star-point on the pre-amp board. Depending on the ground loop situation, tying the shield at one end only can keep 50/60Hz noise currents off the shield, but there are some downsides too, so I'd try both approaches.
Anyway, I haven't done anything with audio in quite some time now, so hopefully someone else will chime in if I'm sounding idiotic.