The only thing I can add to this is that the sample rate of the .txt is just too low to really tell the difference between CW and CCW. Here is an incident where a propulsion issue occurred and the P3S was rotating CCW.
View attachment 92251
This plot was generated from 30 Hz data. The same data sampled at 10 Hz (the .txt file sample rate) produces this plot
View attachment 92252
I can't really tell if this is CW or CCW.
Agreed, but in this case the direction of initial rotation is clearly captured at the 10 Hz sample rate. It then transitions into full 360° rolls, with more minor excursions on the pitch and yaw. It's not quite like the usual spin following motor loss. It's almost as if it lost two motors.