Bit off more than they could chew? Didn't have an extensive background in game building? Perhaps they just never built a 3D game before?
It's kind of like when some of the first 3D games came out, modelers would model a brick then build a wall out of them, instead of just modeling a brick wall other ways, like putting a brick texture on a regular wall. They probably just couldn't come up with a good way to 'cheat' 3D collision detection that worked for them. Approximating is a huge part of game algorithms even with the power of today's hardware.
Because from a distance it actually doesn't seem like so much work. The assets get crazy expensive, but they already had that covered. 3D collision detection and resolving is a solved problem, you can just get off the shelf components (for free), the real problem isn't gettin the game to work, the problem is in the polish.
Jumping over fences, AIs chasing over 3d terrains, making sure people don't get stuck in random stuff. Everything needs weird edge-case stuff that's not in a generic physics engine.
Perhaps they should've first made a 3d game with 2d physics, those games look great and the overhead is much more manageable. But hindsight is 20/20 and from what I've gathered they actually got really close to a great game.
Agreed, game development is the art of making illusions.
Proper simulation, or any kind of complex machinery working behind the scene rarely adds a lot of value to the resulting game (sadly). And moreover, I'm certain that relying of cheap tricks is the greatest virtue in game development. And there is nothing derogative or ignoble about that. Knowing where to cut corners is an extremely valuable skill, and it requires a lot of intelligence, experience, and careful planning.
Computer games are a lot like real-life magic. The best results are achieved if you do cheap tricks, but you do them well..
It's kind of like when some of the first 3D games came out, modelers would model a brick then build a wall out of them, instead of just modeling a brick wall other ways, like putting a brick texture on a regular wall. They probably just couldn't come up with a good way to 'cheat' 3D collision detection that worked for them. Approximating is a huge part of game algorithms even with the power of today's hardware.