Thanks to Pinnacle I have made some new contacts at Xceive and gotten access to the reference drivers and register map for the xc5000. With only a couple of hours of reading and experimentation, I now know why the XREG_BUSY register has been misbehaving, as well as the expected power management behavior. Wow, these things go faster when I can just look at the documentation.
If all goes well, I hope to push out the patch series I had been working on with mkrufky, which will bring tuning time from 3200ms down to around 300ms. Just need to prep the new firmware, write two relatively simple patches, and retest with all the xc5000 devices. Yay.