Natural selection is just an emergent process. Genetic errors at some stage are going to be passed to all descendants, regardless of utility. The harmfulness or usefulness of these errors is left as an exercise for mortality prior to reproduction - which, in modern society, may not be a particular useful factor.
Speaking from my rear, epigenetic influence on growth and development is too fundamental a force to be a random error heretofore untouched by selection. With humans being a social, communal species, adverse conditions tend to hit populations en masse. That means that, for much of our (and our close relative's) evolutionary history, it was relatively certain that, say, all of your neighbors were suffering from the same drought you were. It would thus be "okay" for your children to grow to a smaller, more nutritionally manageable size, because the people they'd be competing with for food, territory and mates would also be smaller. Remember, if genetics are physical destiny, and they say you're going to be big, your body is going to have a go at it, and incur whatever performative costs that come from trying to nourish a 6ft body with a 5'6" diet. Who's more fit: the latter, starving, or the former, relatively well-fed? This model gives the body a mechanism for titrating aspects of development according to local conditions, across the handful or less generations it used to take for those conditions to change.
I would not be surprised if this ultimately explains the global growth spurt that took place in the 20th century.
The popularity of the Darwin awards notwithstanding, mortality is not the only mechanism by which evolutionary pressure is applied - simply influencing the likelihood of having children (and recursively onward through descendants) is sufficient.
That gives wide latitude for more subtle influences (e.g. attractiveness, desire to have children) to have a large impact.
Not the main point, but it should be noted that the Darwin Awards are wide open to candidates with outstanding self-sterilizations, not just people that remove themselves from the gene pool by dying (hilariously).
That's a good point. I just wanted to emphasize that an epigenetic error being passed on is not a measure of its utility to an organism, rather, it is a necessary precondition for the resulting trait to be tested against the environment. Being a necessary element of "natural selection", the meta statement of calling this propagation a positive utility seems a bit incorrect, if only in verbiage.
