Yes. I’ve looked a lot into hitboxes/hurtboxes and experimented with them. More than anything I’d say it’s a matter of the KISS principle. When you make the hitboxes and hurtboxes 1:1 with the model, you end up with values that fluctuate wildly between frames, such as when characters flail their limbs as they are wont to do. This is especially likely to happen when you have a lot of concave and convex surfaces making complex geometries, weird shit is super-prone to happening and you don’t know what the fuck the result will be frequently. Why do you think clipping out of bounds is so common in so many games? It’s usually a result of the hitboxes or the environmental collisions getting distorted so something goes inside something else. When you keep things simpler, the odds of unintended or flat-out weird interactions that nobody can understand go down, especially for collisions where objects affect the position of other objects, like in environmental collisions.
The scale of the hitbox needs to be appropriate to the interaction too, relative to the size of each entity interacting. For example, in Wonderful 101, does each Wonderful one really need a 1:1 accurate model for collision? This also varies depending on the pace of the game, because the fidelity of our perception changes based on the speed of the game, smaller areas that barely make contact are hard to notice, so you can afford to be more forgiving and general in making hitboxes for faster games.
This is why Devil May Cry uses 3 spheres to represent Dante and Nero and no hitboxes on the limbs. Marvel 3 has similarly simplified hitboxes. I’ve seen hitboxes cleanly miss in smash that people swore should hit, and vice versa.
In the opposite extreme, Shadow of the Colossus has extremely detailed hitboxes for its Colossi and extremely simple ones for Wander. The Colossi need complex geometry in proportion to their scale. Wander doesn’t fight enemies around his size, the colossi have much larger implements to hit with, so it only needs to track his central body mass. The Colossi need to be climbed up accurately, the Wander is like a small implement probing a large surface, like the hook dentists use on your teeth. If Wander sticks a limb out and that ends up in your entire body getting crushed then you’ll be shouting about how bullshit that is. You need to be a bit forgiving to make up for difficulties in perception. Also players generally like it better when their hitbox is smaller than represented and an enemy’s is bigger, like in Shmups which took this principle to the extreme. Same for things that the player collects or interacts with, they like it better when their area to interact or collect things that help them is bigger.
In Smash bros there’s this term, Roy-Zoned, for when you’re so close to the opponent that your attacks miss them because they’re on the far side. It’s called that because Roy has a sweet spot close to him, so Roy players try to sweet spot so hard they frequently miss entirely. When someone punches, how much of their arm should actually be the hit area? Just the fist? The entire length of the arm? Some way into the body? Remember, Roy is a character with a sword, he doesn’t exactly have small hitboxes, yet I’ve had silly things like two marths Fsmashing at point blank range and we miss each other. Or try to grab each other and miss. And if you’ve ever seen Marth’s grab box you’d know how silly that is.
It gets sillier when you have the legs given individual hitboxes and someone does like a kick or something where the hitbox fits inbetween the legs, and it’s like, “Seriously? there’s no way that should miss.” I’ve seen this sort of thing happen a LOT in all sorts of games.
You have to think about the hits being the very edges of each hitbox’s range, and would a hit in those regions be acceptable or look weird relative to the animation? Especially true for grab animations, because when you grab the character gets dislocated, teleporting to the held position. The obvious example is Mewtwo’s grab in PM 3.02, because part of the box was inside him, people barely brushing his back would be grabbed and teleported in front of him where he held them. I took a picture of that and have an animated example.
The intention is clear, they wanted crouching characters to go under the hitbox in front so they could get around the grab, but people closer into mewtwo to not get roy-zoned, but it had this unfortunate consequence.
Because environmental collisions in particular are so finnicky, most games don’t use anything close to the character’s model for that. Smash uses a small diamond shape, Dark Souls uses a big pill shape. Mirror’s edge has a cylinder. If you use something more complicated and don’t have a complex animation engine based on inverse kinematics, you’re going to run into trouble there. Even with one you’re going to run into trouble, because GTA and other games like it do.