A modest proposal:

If mothers made a habit of snacking on (each other's) litters of all sons, that would counteract the problem. That wouldn't require being able to differentiate among adult males, just between male and female children.

If the species takes a long time to wean children and doesn't reproduce until that process ends, this works better.

Mice, pigs, rabbits etc. (animals with large litters) already eat weak children fairly often, so this is somewhat plausible.

Rather than rely on females recognizing things about males, what about genes that capitalize on the difference between regular males and those with the disorder - sisters!

Females could more greatly than presently value aggression (this would only need a boost, the trait already exists), and a gene could make females intervene to break up their brothers' fights. Young males with the disorder would tear each other to shreds or be too timid to reproduce, and males without the disorder would have sisters preventing them from killing each other.

There would also be strong selective pressure for any genes which can override the segregation-distorter, even if the females can't recognize the males which carry it.

"Strong" for sure. Unfortunately for the species it would have to emerge fully functional in the time it takes for the species to evolve to extinction. Not so easy.

Nice!

Now I wonder why Eliezer's post calls the original problem unsolved. Surely such elementary solutions couldn't have evaded the experts in the field? I'm guessing that I made a mistake somewhere...