You are right. Indeed I can see it from the first graph of euler.x and at the bottom the trend of quaternion.w (I knew that they can vary only in [-1;1] but I was wrong maybe)
I don't want for now a logic that works in real-time, but what could be at this point the solution to recognize with a threshold the maximum and minimum of the swing with BNO055, in your opinion, please? Do you have some help theory for me?