>The essential difference I see isn't one of individual probability of parenthood, but of aggregate. Not all heterosexual couples rear children, true. But the likelihood of a randomly selected example rearing children is vastly higher than that for a randomly selected homosexual couple. That, I argue, is what makes it worthwhile for the state to support the former and not the latter.

so, according to that principle you declared, the state clearly shouldn't support marriage (no new marriage licenses and no tax benefits to the childless couples with already existing licenses. Shouldn't the state just automatically annul the existing licenses of such couples?) after the age of 50 or 55 - i.e when that likelihood fells below whatever threshold you consider "worthwhile".