What this could be:
This is a good distinction to make, but I think the problem is minor because the player cannot imagine the situation in the game, but the player is drawing dramatically different inferences from the situation, different enough from his that you cannot Effectively plan around them, different enough from the players who find it (involuntarily) chaotic, and different enough from yours (again) that they can be harmful assets.
Here is why I think that:
He knows there is fire around. He caused it. It is part of his character's flaws, so he clearly imagines an environment that lacks something his character wants, fire, which he then creates. But the inference he is not drawing is that it is visible and will attract the monster.
He knows that the monster is too dangerous to fight, so he is also imagining that part of the environment correctly. But the inference that he is not drawing is that aerial night hunters have very good vision to detect movement.
Is could just be simple ignorance on the part of the player. You are in a better position to know, but I am a city boy and I can easily imagine a GM waiting for me to know something about woodwork (say) that I would simply sail directly over my head. That could Being part of the general fallacy of "We have to do something! Running is something! Run for it !!!!"(Everyone dies. Oh, shame).
There could be many things in this general neighborhood that I will summarize as "blind spots."
What i have seen:
I see things like this once in a while and, hell, I've been that guy once or twice. I suspect, but I can't prove, that most people have been that type once or twice; nobody knows everything. So I'm not sure that two sessions are enough to know if this guy is really unusual in his failure to make good inferences, or if he simply has bad luck and is exactly in the wrong situation for him, personally, in the initial scenario.
When I see things like this, about half of them tend to be related to social and social violence ("What do you mean I can't throw a fireball at the thief who got into the crowded market?" Uhh … why will you kill a dozen other people? And maybe you'll burn the whole city? ") And the rest are random things about which a player has really wrong ideas … or at least very different from the DJ.
What I do:
In general, what I do (and I have seen others do) is simply tell the player that his character knows that it is a very bad idea based on any skill or statistic that he has that seems applicable … and that I hope they hear that advice, learn from him and carry it forward in the future. Even if all they learn is: "Wow, I really don't know anything about how aerial night predators operate; I should check with the other players," that can be a victory.
As a GM, you can do this, be the voice of common sense specific to the character's game world. Set up a situation where you feel you are cheating if it happens too often.
How well does it work? It is mixed, honestly. Fireball launchers generally appear quite quickly when they realize that it is not a videogame from the 1980s that will restore everyone's attitudes as soon as they are out of sight. Blind spot characters generally only have to learn to move through blind spots, which is … challenging.
What I advise you to do:
As above, and talk to your player. No confrontations, but try to find out why he was doing those things. And asking is almost always more effective than trying to solve it because, let's face it, you still haven't been able to solve it.
What will not work:
If I am right, and this is just a blind spot or lack of knowledge, then Session 0 will not work. Sessions 0 have the undeclared assumptions that everyone is making informed statements about their preferences and promises, and that they are able to deliver.
But if there really is a blind spot of knowledge, those assumptions are not true, and the player can promise, in good faith, to try to imagine the game world. But it will continue to fail if it lacks fundamentally knowledge about some parts.