As flimsy as it looks, the Odroid once again performed flawlessly, recording 96 minutes of video with data logging, as the vehicle bashed around & flipped over.  

The mane thing it revealed was the sensor was working & the servo was getting the right feedback.  There was no delay between the video & sensor.  The video might have lagged the sensor because of the way the telemetry was buffered.  The video showed the vehicle wasn't responding to feedback.  In some cases, the vehicle responded by turning the opposite direction of the feedback.  It tended to respond less over time, until on the final oscillation, it either didn't respond or it responded by going the opposite direction.

There were no anomalies with the I feedback.  I accum & feedback are in the same units.  What the PID controller deemed to be center is the I accum.  All the cases showed I accum converging on +20.  When feedback goes above I accum, it should turn left.  When it goes below I accum, it should turn right.  There's no obvious correlation between I accum & oscillation, only the P error & oscillation.