Perhaps if we are unable to simulate what should recognizable due to the lack of resolution then at that distance(recognition) the game should use what ever number of pixels are required. The AC will appear larger than it should but at least it would be recognized as it should(a choice of one or the other). The need to have objects larger than in reality would lessen the nearer it got. It would solve the object identification problem but might hinder judgment of closure speeds. But like loosing an eye the brain will soon adjust.
Sorry it seemed simpler to try to explain in my head.