Hah, cool finding. I was considering a situation more similar to "detecting one further photon of light from an already illuminated scene": just-noticeable difference instead of absolute threshold.