Isn't that what epigenetics and methylation-tagging is for? Keep the code for the protein in the genome, but don't express it until some biochemical condition

Isn't that what epigenetics and methylation-tagging is for? Keep the code for the protein in the genome, but don't express it until some biochemical condition is met. Copying long chains of barely-used DNA is relatively cheap for higher-order multicellular organisms with intergenerational intervals spanning 25 years instead of 12 hours. Bacteria are better poised to re-evolve the capability when necessary, but if your species only expects ten thousand generations between solar or geomagnetic events, keeping that anti-radiation gene around for 250 kiloyears makes more sense. Of course, during that time, it may get damaged by mutation and not be repaired by sexual reproduction, but you only need one male with a functional antiradiation gene and enough females to not die before reproducing, and you can repopulate the species within 10 generations of a radiation crisis, instead of spending 100k generations reinventing radiation resistance.