@Quadzilla I just went to look at another one of your logs to try to understand what is happening and the log bitmask is not what I asked you to set again.
Can you please provide logs and summaries of logs all in one place so I don’t have to waste my time getting things second hand.
@Quadzilla, thanks for doing the test. This confirms that it is not an EKF2 vs EKF3 issue which cuts the problem down a lot in size.
My guess (and it is just a guess) is that the issue could be a feedback loop caused by the placement of the autopilot. I think maybe the autopilot is not at the center of rotation of the vehicle (maybe it is higher in the nose of the vehicle?) and the INS_POS_XYZ parameters haven’t been set.
My best course of action is to rebuild this experimental frame but add some extra girth to the arms , possibly move to 960 kv. motors. I will still need to rotate the FC to go the FBWA at some point.