I would argue that dithering is fake resolution, so you're not exceeding the limit in the information theoretic sense. Perhaps a better analogy would be compres

I would argue that dithering is fake resolution, so you're not exceeding the limit in the information theoretic sense. Perhaps a better analogy would be compressed sensing on a sparse, structured signal where you can seemingly violate Nyquist but only because you have additional information about the sampled signal.