bad metagaming is always the GM's fault

Harsh words but I'll stand by them. 

Metagaming is when the player knowledge affects character behavior over and above what character knowledge would afford. Bad metagaming is when we don't want this, which depends on lots of things: most obviously how trad vs. storygamey we want to be, and even in very trad games there will be metagamey elements that arise, most often for the sake of smoothing things out and getting to the interesting bits:

• leaning in to cooperation with other player characters, even when the characters would have more reason to conflict
• speeding up/slowing down action because you know the IRL clock on the game session is going to end
• choosing to steal the jewels at night rather than by winning over the prince at the ball because you (the player) like heists more than social adventures

or whatever. [1]

The classical instance of bad metagaming is something like: the party are attacked by trolls, but the players have cleverly read the Monster Manual and know that trolls can't regenerate from fire, and so break out flame even though their characters are benighted innocents who are unaware of such things. [2] Players, don't beat yourself up over this; GMs, this scenario is your fault and you should have got gud. This is on you because:

1. Players don't have control over whether their own knowledge propagates into character action.
2. You DO have control over the fictional situation, and hence whether player knowledge corresponds to it. 

Within a trad or simulationist context metagaming tends to be bad because it takes people out of actor stance, but when players are "playing dumb," the damage is already done. You are the one who decided that trolls die from fire and acid damage, not the Monster Manual. The Monster Manual just suggested it to you, just like it suggested it to your players. 

Does fire and acid damage route around troll regeneration? Do trolls regenerate at all? Make those decisions in the context of the fact that the players have been suggested that this is so. The easiest assumption to make, the one that requires the least work from you and players, is that meme D&D (or whatever folk knowledge associated with the kinds of books players know you're borrowing from) is within the folk knowledge of the setting. Communicate that this is the case and that this folk knowledge is in many cases correct and in some cases incorrect. Then you can decide whether trolls regenerate and what they don't regenerate against, and when PCs encounter trolls (or creatures described similarly to them) they can start with flame as a starting hypothesis and proceed empirically from there, experiencing neither dissonance nor betrayal regardless of the result.

[1] I think most of these kinds of good trad metagaming are best thought of as costs rather than good in themselves, insofar as they take you out of an intended actor stance and into author stance, as opposed to the kind of metagaming one can do from author stance, like deciding that your character who has decided to seduce the prince also finds him absolutely revolting because you (the player) think that would be dramatic and funny.

[2] Maybe this scenario is good at your table; substitute another scenario, then.