Flyaway (indoor !) need help to analyze log

I need help because i can’t find the reason for this failure in the log:

This was an indoor flight in PosHold mode, using an optical flow sensor and a rangefinder.
The copter did not keep height well, so i reduced throttle 3 times. Suddenly the copter climbed until it crashed at the ceiling, although i had reduced throttle to min.

Can anyone tell me what went wrong ? (in my view, all sensor readings where good)

Here’s the log: (673.8 KB)

EKF estimation failure, possibly due to accelerometer aliasing.

Consider sticking some mass to your autopilot to see if that stops the aliasing.

Also consider setting your EKF altitude source to be rangefinder - if you’re only going to be flying inside over a flat plane.

Thank you @peterbarker for the explanation which is probably correct, athought i don’t see any reason for such high frequency vibration that could lead to sensor aliasing. I’m flying this copter for years now and never had such issue before 4.1.3. This is not a racer, but a 450 quad, 9.5 inch props.

One question remains:
There exist two Imu’s, so there are two EKF “lanes”. The estimation of lane No.1 is quite good according to XKF[1].PD, the estimation of lane #0 is obviously ruined. Why did the firmware choose the wrong lane, given the good baro readings ?