A new solution to the non-identity problem?

Molly Gardner in a piece that just came out offers an interesting new solution to the non-identity problem, the problem of making sense of benefits and harms to people who wouldn't exist were it not for our actions of benefiting or harming. Gardner's suggestion is:

A state of affairs, A, is a benefit for an individual, S, just in case if it were true that both S existed and A did not obtain, then S would be worse off in some respect.

This is a clever solution: normally when evaluating whether an action benefits or harms someone, we simply ask how they would have done had we not done the action; but Gardner wants us further to keep fixed that the patient exists.

But clever as it is, it looks to me that it fails. First, suppose that a strong essentiality of origins thesis obtains. Then whenever we benefit or harm a future person, that person couldn't exist without our action. But that means that Gardner's conditional becomes a per impossibile conditional. And concepts of ethical importance should not be defined in terms of something as poorly understood and as controversial as counterpossible conditionals.

Suppose now that there is no strong essentiality of origins thesis. Then, plausibly, a person who was conceived through coitus could also have been conceived through IVF, at least if the same sperm and egg were involved. Now suppose that where the couple lives, IVF technology is highly experimental and works so poorly that children conceived through IVF end up having all sorts of nasty health problems. The couple is wicked and doesn't care about the health of their children, but they also haven't even heard of IVF, and so they conceive Sally the natural way. Now let's consider Gardner's conditional. What would have happened had the child existed and the couple not engaged in coitus? Well, the closest possible worlds where Sally exists and the couple did not have intercourse are worlds where the couple engaged in a poorly-functioning IVF treatment, and hence worlds where Sally has nasty health problems. So the couple benefited Sally by engaging in coitus.

The conclusion that the couple benefited Sally by coitus is, I think, true. For I believe it is always good to exist. But it is clear that Gardner doesn't want to suppose that existence is always a good. And if existence is not always a good, then we can suppose a scenario like this: Sally is going to have an on-balance bad life if she is conceived by coitus, and an on-balance worse life if she is conceived by IVF. By Gardner's criterion, the couple has benefited Sally through coitus, even though Sally's life is on-balance bad. This is surely mistaken. One might say that the couple benefited Sally by engaging in coitus rather than IVF. But since they never even considered IVF, one can't conclude that they benefited Sally simpliciter. (If Sam gives Jim a mild electric shock, he harms Jim simpliciter, but he benefits Jim by giving him a mild rather than severe shock.)

And even if we grant--as in the end we should--that existence is always good, Gardner's conditional gives us the wrong reason for thinking that the couple benefited Sally. For the benefit to Sally has nothing to do with the fact that Sally would have been worse off in the nearby worlds where she existed through IVF.

And even if essentiality of origins is true, the argument concerning Sally works. For it is still true that, per impossibile, had Sally existed but without her parents having intercourse, she would have existed through IVF and hence had very poor health.

The problem with Gardner's approach is this: the worlds that are relevant to the evaluation of her counterfactual may simply be irrelevant to the question of benefit or harm simpliciter.