Which means that antinatalism/negative preference utilitarianism would be willing to inflict massive suffering on existing people to prevent the birth of one person who would have a better life than anyone on Earth has ever had up to this point, but still die with a lot of unfulfilled desires.

Is that really how preference utilitarianism works? I'm very unfamiliar with it, but intuitively I would have assumed that the preferences in question wouldn't be all the preferences that the agent's value system could logically be thought to imply, but rather something like the consciously held goals at some given moment. Otherwise total preference utilitarianism would seem to reduce to negative preference utilitarianism as well, since presumably the unsatisfied preferences would always outnumber the satisfied ones.

Yes, but from a preference utilitarian standpoint it doesn't need to actually be possible to live forever. It just has to be something that you want.

I'm confused. How is wanting to live forever in a situation where you don't think that living forever is possible, different from any other unsatisfiable preference?

If the disutility they assign to having children is big enough they should still spend every waking hour doing something about it. What if some maniac kidnaps them and forces them to have a child? The odds of that happening are incredibly small, but they certainly aren't zero. If they really assign such a giant negative to having a child they should try to guard even against tiny possibilities like that.

That doesn't sound right. The disutility is huge, yes, but the probability is so low that focusing your efforts on practically anything with a non-negligible chance of preventing further births would be expected to prevent many times more disutility. Like supporting projects aimed at promoting family planning and contraception in developing countries, pro-choice policies and attitudes in your own country, rape prevention efforts to the extent that you think rape causes unwanted pregnancies that are nonetheless carried to term, anti-natalism in general (if you think you can do it in a way that avoids the PR disaster for NU in general), even general economic growth if you believe that the connection between richer countries and smaller families is a causal and linear one. Worrying about vanishingly low-probability scenarios, when that worry takes up cognitive cycles and thus reduces your chances of doing things that could have an even bigger impact, does not maximize expected utility.

I think, for instance, that a total utilitarian could at least say something like "no less than a thousand rush hour frustrations, no more than a million."

I don't know. At least I personally find it very difficult to compare experiences of such differing magnitudes. Someone could come up with a number, but that feels like trying to play baseball with verbal probabilities - the number that they name might not have anything to do with what they'd actually choose in that situation.