Let me, as often, think out loud.
Counterfactuals seem to be an extremely powerful philosophical tool. Except that they never seem to work. Think of the employment of counterfactuals in connection with epistemology (e.g., one only knows p if one would not have believed p had p not been true), the theory of truth (Russell: p is true iff to believe p would be to believe truly), causation (Lewis), free will (e.g., x does A freely only if x would not have done A had x not wanted to do A), intention (e.g., x does A in order to achieve F only if x would not have done A had she not believed that it would achieve F), etc. It seems largely, and perhaps entirely, a history of failure. Yet, at the same time, counterfactual claims continue to seem tantalizingly close to capturing something important about many phenomena. Counterfactual characterizations are roughly right, but then fall apart when the details are to be worked out, or odd cases are considered (sometimes, the initial counterexamples are odd indeed, but with more work one can see that these counterexamples are not quite as out of the way as one might have thought). Counterfactual accounts are roughly right, but they cannot be modified to be exactly right. These are surprising facts, and it would be nice if a theory of counterfactuals explained them.
A standard story is that in a lot of the cases where counterfactual relations aren't doing their job, it is because one also needs an "in the right way" constraint on the counterfactual. Now, taking the words "in the right way" literally suggests that a normative, proper-function based, constraint is needed.
Could there, perhaps, be something like a counterfactual but which has that constraint built-in? But the force of that constraint is different in different employments of counterfactuals.