I'm not sure if anyone else has run into this problem, but after a lot of work to eliminate any potential ground-loop issues, it seems that having the brake enabled in an electronic speed control can destroy the new power module with built-in BEC, which in turn damages the ArduPilot board. We destroyed two sets before isolating the problem to the ESC brake.
We have these mounted in RiteWing Zephyr IIs, and we're using a Hacker A30-10L V2 and a Castle Creations Ice-Lite 50, with a peak static load of about 47Amps. We are not using the BEC in the ESC, only connecting its signal line to the autopilot. We use a CC-BEC soldered directly into the power module to feed the servo power, and the power module runs the APM with it's built-in BEC. The jumper next to the servo output headers is removed to separate the two power feeds.
With braking enabled, a few quick pulses to throttle will immediately knock out electrical power. Looking carefully at the power modules, there is a chip on the BEC daughter board that appears to be blistered, and of course the module reeks of smoke. Some of the LEDs light up on the autopilot, but it no longer communicates with anything and appears to be dead - never completing a boot and never passing RC signals through. The 3DR radio, GPS module, Pitot module and RC receiver all seem fine, even after two of these "events."We managed to use the brake without a problem when there was no prop on the motor, but this means low angular inertia, so not too much regenerative power when the brake comes on. With a prop, the built-in data-logger in the Ice-Lite records a voltage surge while braking that is roughly double the battery voltage, so equivalent o a 6S battery. The duration of the surge is very small, a fraction of a second, but it seems that this voltage leaks back to the power module, and something there can't handle the brief voltage spike.
We've tried this with another brand ESC that has weaker braking, and it seems that the power module and autopilot can survive this.