Every time I was enticed to look up more about hormesis, I see the same issue: an intrinsically linear effect of a factor is studied, with a precise linear generator of the factor, but outside of the generator there is a background component to the factor which is ignored, which causes a misinterpretation of non-linearity.

A concrete example, suppose you have a light sensor in a "dark" enclosure, then at large enough intensities the current through the photodiode is linear with the incident illumination, but if there is some light leaking (or alternatively thermal radiation, and hence temperature, and hence dark current) then as the light generator is set to lower and lower levels, the light sensor will no longer linearly approach 0, since the signal starts to delve below the noise floor (so it will allways be measurable, but require more and more oversampling to decrease the noise floor). To confuse this effect which has nothing to do with photons getting converted to electron hole pairs, it is a misinterpretation to consider the effect "non-linear" close to the noise floor, and an even bigger misinterpretation to consider it "beneficial". Sure even high levels of ionizing radiation can be beneficial to the offspring of a colony of bacteria, fungi, or plants as a group, but it most certainly is harmful to the the individual bacteria, fungi or plants individually.

In the case of the light sensor, the current through the reverse biased photodiode will still be ideally linear with the total incident illumination, just no longer linear with the illumination of the non-dominant light source.