I did some testing recently with helicopters trying to solve loss of control issues with a EKF switch on these v3 controllers where IMU’s never seem to agree. I first did several days of simulated testing on the ground and figured out how to reliably produce a GPS glitch and subsequent rapid switching of the EKF processes. I thought I had it fixed by setting INS_USE2&3 to zero. So I tested in flight. I experienced total loss of control of the helicopter by initiating a GPS glitch on-purpose. The second attitude solution is totally different from the first, then it quickly switches again and another attitude solution that is different, it keeps playing musical chairs and the heli goes totally nuts. The only way to recover it (with helicopter) is to have flybar mode turned on and fly the aircraft no-gyro in Acro flight mode once the EKF goes bonkers.
The “fix” here was to set EK2_IMU_MASK to 1. It can then handle a GPS glitch and nothing bad happens. Or replace the controller with a decent one that has IMU’s that are consistent. The code allows up to 1 m/s/s difference on accelerometers before it will flag it on the pre-arm checks. But unless they are identical none of the supposed “redundancy” works when the EKF starts to rapidly switch “lanes”. You’re better off to just use one IMU, set up your system for pass-thru control that does not depend on any of this, and rely on your pilot skills for redundancy.