Whatever the CMOS pixel size, whatever the image processing or stacking.
Dura Lex Difractionis, Sed Lex Difractionis ==> If you can "model" (i.e. assume) a telescope on Earth with an aperture D (in meters), the resolution of your telescope CANNOT be better than lambda/D in radians, with lambda = 500nm (mid of visible spectrum). If you can play with interferometry, the largest baseline of your interferometer (maybe 100s of km!) is D (but it's difficult in visible light). Any other assumption is a "fictional" telescope.
Dura Lex Difractionis, Sed Lex Difractionis ==> If you can "model" (i.e. assume) a telescope on Earth with an aperture D (in meters), the resolution of your telescope CANNOT be better than lambda/D in radians, with lambda = 500nm (mid of visible spectrum). If you can play with interferometry, the largest baseline of your interferometer (maybe 100s of km!) is D (but it's difficult in visible light). Any other assumption is a "fictional" telescope.