Isn't it well-known that Nash equilibria of games are highly sensitive to the modelling? And that seemingly benign modifications can lead to much worse outcomes?

Eg Braess' paradox, Induced traffic/demand, strategic voting, etc.?