+1 for giving a thoughtful response.
In war games, we call the players that game the rules, rules lawyers, vs metagaming. Could you call it metagaming? Sure, and that puts you right back into my point is that there are acceptable and unacceptable levels of metagaming in RPG's. Pretty much the same with war and board games where there are acceptable and unacceptable levels of rules lawyering. The fact that you are mentioning as a catch 22, which it is, I agree; it is a lie agreed upon. No metagaming except what is considered acceptable. Fate points are a good example of the whole argument, vs other games, because that's the target definition moving for each game. It's like the party composition, because that could be thought of as playing the game also, yet it is still metagaming, it's something you need to do to make the game function.
Okay, let's try again. In a wargame, yes, you can exploit rules and that's likely metagaming (gaming the game), but you can also know your opponent, and know that when Bob plays, he likes to flank left, so you can set up a feint to draw in Bob's flanking left and then crush it. This is also metagaming; you're using your out of game knowledge about how Bob plays to beat him in the game. Normally, this isn't considered bad at all, but part of the competitive nature of wargames.
Metagaming just means outside of the game. It's what the word literally means, "meta" meaning "beyond" and "game" meaning, well, game. Metagame concepts exist outside of the defined scope of the game. So, if there's a rule in the game, like Fate points, then that is not metagaming because that's not outside the game.
In RPGs, this gets twisted up a bit, as there's a large swath of players that consider thinking only in character and only with the information the GM provides to the character to be a constraint on players in the game. In this case, knowledge of trolls and fire may be metagaming if that information has not been afforded to the characters by the GM yet. The catch-22 comes from the fact that you can't actually remove the knowledge from the players, so they're acting on out of game knowledge one way or the other. It's generally preferable by groups that play in this style to pretend ignorance until the necessary gates are opened by the GM, but that doesn't actually make the play non-metagame. It's just the preferred metagame.
Quite often, metagame is used as a synonym for cheating in these discussions, rightly or no. What's missed by this is that metagaming is part and parcel of most games, especially the social ones. You're always tailoring your play to how your GM and other players engage the game. If, for instance, you know your GM doesn't allow hiding in combat (a seemingly not-uncommon ruling in 5e) so you don't spend resources on stealth because it won't come up often, then that's metagaming -- you're playing the game based on out of game considerations. I don't think anyone would be too upset by this, though. However, in the metagaming = cheating mindset, this is often just flatly denied to even be metagaming, despite it clearly being so, because metagaming is only those behaviors that are disapproved. This is the problem with frank and clear discussions about metagaming -- it's far too loaded a term in RPG contexts because how it's defined is so variable based on the individual.
This conversation we're having is a good example of these variable definitions -- you keep moving back to metagaming being things that don't obviously follow from the fiction (Fate points) with a side of the "secret" knowledge wall (trolls and fire). I'm using metagaming as it's actually defined and not considering it automatically negative (in fact, it's often beneficial). As such, we've been unable to effectively discuss metagaming.