Rev 1 of the PCBs arrived and seem to work great with one caveat.
I built up two boards. Both work, running the existing code, but one streams at a slower rate than the other. With this design I moved to modules using the v3 ESP32 chip and even after recompiling for that version the slow-down on one module persisted. I have noticed that one can get the FreeRTOS scheduler stuck in a pattern and I wonder if something like that is happening. Or perhaps something with one module's wifi interface. I'm going to add debugging code to see what part of the code is slowing down on one module.
The board has the ability for the ESP32 to reset the Lepton and I'll add code for that and do some clean-up. Then I should be able to post the code and hardware design on github.
Probably also eventually rev the board to add a reset button (useful to restart while the serial monitor is attached) and tighten up the cutout for the LED to make assembly easier (or hopefully find a cheap side-looker RG SMT LED with lens to poke through a 3D printed enclosure (also have to design the enclosure...)