Revision A of v3 of the board has gone out in an order. Hopefully they arrive soon so that I can post an update and show a working board. In the meantime...

There were a number of last minute hacks I had to make. I didn't notice this, but the AZ1117-3.3 regulator I had selected originally was pretty unusual in that it placed the output rail on pin 3 instead of thoug the tab (SOT223-tab) and pin 2. Result:

I'm visibly disgusted too, but I wanted the order out of the way, and the components were the largest cost overhead.

Another gaff was running the DI pin trace for the RGB LED, all the way... across the other layer, all the way through the ground plane. It shouldn't cause any issues, it's just inelegant and probably a result of rushing the component layout.