If I understand Erik right, the problem with the oscillations is not their presence but frequency that can effectively lock our program in interrupt read routine and never allow normal program loop to run, plus corrupt our summary value if we do it in some silly way, without taking previous readout value into consideration.

The transitions even though they are 0-1-0-1-0... are so high freq that we will miss readouts of most of them and should consider these we read as completely random string of zeros and ones, so without a sane transition recognition, our "angle value" will float randomly all around. If we run on incoming new value driven interrupts too, and don't limit their rate, we will never get to do anything other than reading the new incoming values.

Is this the right summary of the problem, Erik?