After thinking about it a lot, I think Might is kinda non-intuitive, but not for the same reasons as some other posters - the problem, in my mind, has more to do with Dexterity and Constitution than the difference between Might vs. Strength.
As a player, I'm used to seeing Str, Dex, and Con as a set. Sometimes Dex will by Agility, or Con will be Vitality, or whatever - but the three are usually seen as a set of "physical" stats, often with 2-3 mental stats accompanying them. So when I see Might, Dexterity, and Constitution coupled with Intelligence, Perception, and Resolve, it feels familiar, and I reflexively start looking for how the six stats go about playing to their archetypal roles. When they fail to do that, because Might really isn't Stength, it can be jolting, like a cognitive "skip" in a record. I may want to ignore it (not sure); I may like the new status quo (I do); but it still forces me to adjust. And I don't think this is a unique experience.
I get that the game designers don't necessarily intend me to, and I understand that I should approach PoE as something fresh, but that is what it is. I have some limited experience with tabletop RPG design, and what I've learned, overwhelmingly, is that even things which make perfect sense can be problematic. For instance, my group had a system where any stat could be used to attack or defend (subject to class limitations), which led many people to question how that could be. To us, and to most of the people we explained it to, this made perfect sense - but it still put off some prospective players, even ones who thought our reasoning made sense. When something is non-intutive, even for really objectively silly reasons, it can break verisimilitude, or the feeling of realism (which is far more important than actual realism).
Personally, I'd be inclined to either repurpose the Might stat to be more like Strength (keeping the name or otherwise), or to at least change the names of Dexterity and Constitution (possibly to something more abstract and more "mental" in tone). Either would allow things to be more intuitive.
Hmm, but why can't Might be interpreted as Strength? This can be a magical system where your muscles provide a component of the magical energy. You're "throwing" a magical force at something, which gets amplified through the spell. That approach doesn't seem any less implausible to me; it's just different than D&D. Now, a wimpy, intelligent wizard has to rely on attrition tactics to win his battles, rather than, say, delivering a knockout blow against the enemy leader.
I don't think it's that there's anything inherently wrong with that approach, but for a certain number of people it just doesn't feel "right," so to speak, for their Medieval European fantasy. You could give all kinds of reasons - it breaks form, it feels "too anime" ( ), it's not what they felt they were signing up for, etc. I'm not 100% sure of where I stand on this, myself.
On the one hand is the universe and its internal logic - it does make sense with PoE's notion of "spirit" as the power source of all characters. Fighters and barbarians and so on are ultimately a kind of mage in PoE, so it's not much of a leap to think that wizards and cyphers might benefit from well-conditioned bodies just as their counterparts do. I love this kind of internal consistency, and the sense of a very cohesive world that can often come with it.
But on the other hand, making these kinds of radical shifts away from what's expected seems to be at odds with having such archetypal divisions as "fighter," "wizard," "cleric," and so on and so forth. I get that PoE wants to simultaneously invoke a feeling of the familiar while also going its own way, but when people hear "wizard," and they have certain expectations for what that means, especially when they hear it right next to other terms that it's often paired with. Veering away from those expectations can be offputting, even for someone (like me), who really wants to like the direction things have gone in.
Or you could accept that people do not find this system very intuitive and would like a more "simulationist" approach where they feel like the derived attributes really fit to the attribute that governs them.
In fairness, the current approach is no more or less simulationist than the alternative - just different in a way that can be jarring.