Doable for a single bullet, yeah. But once you add things like miniguns, flamethrower, etc. the calculations can get out of hand easily.
Meh, you don't have to have some sort of flame physics. A flamethrower would essentially be an arc of damage which should offer, from an implementation point-of-view, a challenge no bigger than field-of-view. Minigun? Just a bunch of bullets. I think you overestimate how costly it is to calculate a bullet's flight path. Shooters have been able to calculate dozens if not hundreds of them simultaneously in real-time since eons ago after all.
Don't get me wrong, from a pure dev and tech perspective, I'd love a full simulation
But as a dev myself, I understand very well why this isn't done anymore since 3D environment became more... detailed.
Not done anymore? Pretty much all shooters simulate the bullet's flight path though? The reason they don't do it in strategy games like these is probably because of a combination of laziness and inability to see the value of it.
Real-time games have entirely different demands, especially shooters. Obviously, I wasn't talking about those
In a shooter, you do not need to simulate the result of an action beforehand to give the player a summary if he should pull the trigger or not
Instead, the simulation advances in real-time, just as the game does.
Also, shooters only rarely feature realistic hit zones, instead relying on abstracted hit boxes in order not to calculate everything. And that is true not only for characters, but the entire environment. How often did your shots get stuck in air because of this?
This is an issue with FPS games to this date - though admittedly it did improve over the years with the CPUs and GPUs improving.
There is - to my knowledge - no game that does a complete physics simulation not relying on simplified abstractions (or being abstract to begin with, like Minecraft).
Another point is physics. There is simply no need to feature anything more than a very basic physical simulation (for effects, mostly) if you do not want a proper ballistic simulation.
But as soon as you need that, it will cost a lot of effort (after all, things can bounce off, so devs need to have different attributes for different materials, just as one example of many) and just having the simulation running all the time costs resources as well. Resources that may take away from AI.
To do it properly involves much, much more than you may think. There is big chain of "If we have this, we also need this. And if we have that, we also need...".
And if you did not do it properly, it would just suck. A strange bug in the physics part of the engine? Could throw you back weeks or months. A bad simulation is much worse than no simulation at all.
Now I'm not saying it's impossible. Of course not. But would each toaster be able to run it?
Not yet, anyway. And games like this aren't usually aimed at the high-end-PC crowd.
And, again, there are far more important things than the few % such an approach would yield over the much easier one based on % rolls.
