I'm working with a ADIS16480 to test attitude estimation algorithms.
I have noticed that, trying to acquire sensors measurements in the body frame and attitude estimation from the IMU, the informations from accelerometers, gyros and magnetometers are easily obtained.
Instead the attitude data, during IMU maneuvering, sometime suffer a sudden block for some second (I've not measured how many).
For the data acquisition I used an Arduino Due board and also the EVAL-ADISZ board.
Reading the ADIS16480 data sheet, I found that this could be caused by EKF divergence.
I would ask if, for you, this could be the problem or it could be something worse.
Also I would ask informations, if possible, about the Linear Acceleration/Magnetic Disturbance detection parameters (page 29 of the data sheet), how they work and they are defined and if they might affect this strange behaviour.