We should improve it by not dropping the audio, but using smooth resampling ratio changes instead. We need to re-read chapter 1 of “Audio Anecdotes” that is about human-detectable pitch changes, to tune the speed at which we want to correct the drift. We also want to re-read the paper "Controlling adaptive resampling" by Fons ADRIAENSEN (http://kokkinizita.linuxaudio.org/papers/adapt-resamp.pdf).

We can reevaluate the ranking on this one, based on how well the dropping works in practice.