I think the premise of a „pixel“ being the smallest entity in software is not right. Rasterization, i.e. translating (actually reducing) a defined subset of the software state into a 2D grid of colored pixels, is only a very limited view on that software.
This might be the reason for the different answers we‘re getting here. Most aim for subatomic physics, it we could also go to light theory (photons and wave frequencies/resolution) and human retinas, general optics and electron microscopes, which again would end up at subatomic physics (you got my circle-train of thought here).