A quick thought: I previously characterized metaphysical possibility by asking about what worlds "had the opportunity to be actual". But this 'positive' frame leads me to have quite restrictive intuitions, even to the point of wondering whether any other worlds had a genuine chance of becoming actual in place of this one. We might instead frame the issue in terms of the 'negative' question: is there any reason why world X couldn't have been actualized? And now I'm more inclined to the more liberal (and standard) view that anything conceptually coherent is allowed. So I'm inclined to different answers. But are they different questions?
If a world never had any opportunity to be actualized, then that presumably means that its actualization couldn't have happened. Both phrases are saying the same thing. But we shift the default presumption by asking 'why?' rather than 'why not?'. Given that these fundamental questions are so difficult and confusing, I'm inclined to shrug my shoulders sceptically either way. "Did world W have the opportunity to be actual?" *shrug*, not that I can tell. "Is there any reason why W couldn't have been actualized?" *shrug*, again, (if W is a coherent scenario) I'm not aware of any such reasons.
Should we prefer one of the questions over the other? Perhaps the negative one is preferable, since it doesn't appeal to the notion of "opportunity", which -- as previously noted -- becomes awfully slippery when abstracted from a concrete causal and temporal framework. (Can we speak meaningfully of atemporal chances or decisions, or of time being caused to begin?)
In addition to allowing us more confidence in the concept's coherence, the negative conception might also provide us with a better epistemic grasp of the facts in question. Though this will depend on our answer to the question: what sorts of reasons might disqualify a world from possibly being actualized? Logical incoherence is one obvious answer, and I can't imagine why there would be any others. But we can go two ways here. We can take it as primitive (a "brute modal fact") that some worlds get disqualified and not others; then the epistemic problems remain. But we might instead understand these modal facts as being grounded on further reasons -- and if we accept my above answer in particular, we end up with a form of modal rationalism, whereby modal notions are constitutively tied to rational notions of a priori coherence and the like. This seems like a more productive line to take.
We might support this conclusion by way of thought experiment. Suppose it is a brutely necessary fact that God (Creator of worlds) is averse to unicorns. Then, as a matter of free choice, God would never choose to actualize a world with unicorns in it. You can "rewind and replay" the creation of worlds as often as you like, and you'll never end up with any unicorns. It seems that by my original definition, then, it's "really necessary" that unicorns do not exist. But, intuitively, that seems the wrong answer to give in this sort of case. God could (really!) have made a unicorn world, if he'd wanted to. He just always chooses not to. But that's no fault of the unicorns; there's nothing about the intrinsic nature of unicorn worlds that would prevent them from being actualized. So we should conclude that they are possible in the broadest sense.
Oddly enough, "brute modality" doesn't seem to capture this intuition, at least not on the "rewind and replay" conception. We might get around this by finding a different conception of brute modality, which allows us to say that God brutely could have chosen differently, even though he absolutely never would, no matter how many times you got him to replay the decision. Or we might allow some non-primitive elements to inform our modal analysis, perhaps tying it to rational notions as discussed above. I think I'm becoming more sympathetic to the latter.