== Edit 2/15 ==
I've replaced the 2.9 APM_RC libraries with the libs from 2.8, and commented out some code in radio.pde that referenced those libraries. The issue's disappeared using this workaround so at least I can fly on 2.9 using PPM.
I'm running 2.9 on a Crius V1.1 board with HobbyKing 30A SimonK (bs_nfet.hex) firmware. Motors are NTM 2826 1200kv motors with 8x4 carbon props. Receiver is a TFR4-B running in PPM mode.
This is all on a tricopter with an AUW of 1180g.
As the video shows - there are random power surges in the front 2 motors. For some reason the tail motor doesn't see these surges. I've verified that both front motors and escs are working properly by hooking them up directly to the receiver, so solder connections are all solid. I only see the power surges when it's hooked up to the flight controller (could there be a problem with the header pins?)
At the instruction of Aleksey - I played around with RC_SPEED which defaulted to 490hz. I tried 400hz and 125hz. Lowering the refresh rate to 125hz actually made the defect larger. It surged harder and longer before it caught itself and settled down to the right speed.
At this point - the only thing I can think to do is swap out flight controllers - but that's going to take some time. I thought I'd throw this out there to see if anyone's ever seen anything like this.
I'm attaching tlogs from both a flight and the most recent bench run. My board can't do raw sensor captures, so all I have is whatever standard telemetry gives me. All I can tell from the logs is that channels 1 and 4 go high and low at random intervals.