Cass Sunstein wrote a great review lately of a book about counterfactuals. The point I thought was most important was the last one:

Evans is aware of the claim that 
whenever historians designate a cause as necessary, they are implicitly using counter-
factuals. In his view, however, “normally historians are not so bold as this,” and they prefer to use the word “probably.” But this response is hardly sufficient. If historians say that one factor was “probably” a necessary condition, they are engaging in counterfactual thinking, just with a qualification. (Surely Evans would not withdraw his objections if Ferguson, Charmley, and other counterfactualists had inserted “probably” in the appropriate places.) Evans adds that even when historians call a “cause necessary rather than possible or contributory, they almost never speculate about the alternative course events might have taken had it not been operative.” This misses the point. Whenever historians call a cause necessary, they are, in the same breath and by virtue of that very statement, speculating about an alternative course.

Beyond that, I will say that I have gotten a lot of value out of counterfactual reasoning, but I only express it on my blog, not my actual "academic scholarship." For me, counterfactual history is useful only if it helps highlight something clever about the past that you might not have seen otherwise. For example, I once wrote about whether you could have an atomic bomb in 1945 if Albert Einstein had never been born. I concluded that you probably could, which was really just a sneaky way to point out that despite the firm connection in people's minds between E=mc^2 and nuclear weapons, it was actually a pretty tenuous one, and that the scientific work that led to nuclear weapons was actually somewhat separate from the kind of physics Einstein was working on. Another post recently discussed what would have happened if the first atomic bomb had been dropped on Tokyo in March 1945, instead of the Tokyo firebombing raid. This was really just an opportunity to explore the deadliness of firebombing versus atomic bombing, and to point out (which comes as a surprise to many people), that even if the Tokyo firebombing raid killed more people than the bombing of Hiroshima, an atomic bomb on Tokyo would be expected to kill several times more people than the Tokyo firebombing raid, because the atomic bombs were far more deadly (in terms of expected mortality on a given population size) than the firebombs.

So this is all a pretty conservative version of counterfactual history — not trying to spin out some long, "what would the 21st century look like if the Nazis had won World War II" sort of thing that gets so far outside of historical facts that it just becomes a work of pure fiction. Instead I'm using counterfactual reasoning to try and play with one or two variables, as a means of sussing out (per Sunstein) whether the variables are as "important" as we typically describe them as being.

I heard Morstein-Marx give a talk on whether Caesar had to cross the Rubicon a while ago. I'm not sure it's strictly counterfactual ("what if Alexander marched west?" I think is the most common kind of question I think most people mean by counterfactual history), but it's an interesting use asking about other possibilities to arrive at new conclusions. His method was to employ game theory - a kind of adaption of The Prisoner's Dilemma. I don't remember all the details, I'm afraid

The fact is Caesar crossed the Rubicon, and war broke out between him and Pompey. Was this inevitable (as some biographers say)? What are the other possibilities? Caesar might have surrendered and returned to Rome. Caesar might have stayed in Cisalpine Gaul. Caesar might have crossed the Rubicon and reached a negotiated settlement. The first two possibilities are clearly not winning conditions for Caesar, but the third, the co-operation choice in the terms of the prisoner's dilemma, is clearly the best outcome; it avoids the zero-sum result of playing the dilemma in the traditional "defection" way.

Morstein-Marx noticed that as late as January 25, Cicero is writing letters to Atticus expressing hope for a settlement, almost a month after Caesar crossed the Rubicon, so that act does not seem to have been the point of no return it has become in history (partly because, the argument runs, no source until Lucan mentions Caesar in connection with the Rubicon. I have problems with this interpretation, but it's not my argument).

So, if co-operation was still a feasible outcome, what forced the players to choose defection, and which player defected? Caesar says he was constantly sending letters asking for a meeting to settle things. Pompey has plenty of time to leave for Greece, but only does so at the last minute, so it doesn't appear either side wanted to defect. However, the consuls, who had rejected Caesar's earlier offers, had already left for Greece, and apparently Pompey didn't feel competent to negotiate in their absence, after all he's officially just governor of Spain at the moment. Both consuls were part of the hard-line minority that earlier had prevented a compromise the majority of the Senate had agreed on (that's Appian BC 2.30-31. The vote was 22 against, 370 for). The hardliners effectively prevented either Pompey or Caesar from playing co-operation.

At least that's the argument I remember. Considering other possible outcomes allows you to determine whether the historical fact was inevitable, and to locate cause in different places. This argument is interesting because it places blame neither on Caesar nor Pompey nor the collapse of the Triumvirate, but on a hardline minority. Whether I buy this argument is a different matter.