Hi pizza,
(1) If you don't position your IMU in the center of gravity of your plane, a rotation of the plane around it's center of gravity (eg a roll from 0° to +45°) will cause an acceleration on your IMU (because it changes position compared to the earth, and position changes can only occur when it has a certain speed/speed change).
(2) A high sampling/processing rate is always better because
(a) quick updates means very small changes in angles which can better be threated as if they would be linear (cos(0.000001) ~ 1). Imagine you update only once a second. During that second, the plane pitches 90° and after that it rolls 45° (use your hand to visualise this
). In this case, you receive the sensor readings, but you don't know whether it pitched first, or it rolled first. This means you can never make a correct estimate.
(b) your ADC sampling rate is better always at a higher frequency than your noise (2 times higher - nyquist - in theory, 5 times higher in practice) so you can average it out, otherwise you could always measure on a noise spike (assuming you get an instant reading). In practice, most ADC's give you an average reading over the sampling time, so this is less of an issue.
Tom