Depends on how anal you want to get about your sim... using lookup tables means you have to have a lookup table for every situation you want to simulate.
I'd say you'd have a table for every condition you feel is important to replicating the fundamentals of the problem.
So a basic lookup table is just air resistance at x speed.
Unless Cd is showing a dependency on speed, it's no lookup. Just Fdrag=rho*Cd*A*speed^2/2, Cd, rho, A being fixed.
Want to include drafting? Second table, including if a car is in front of you.
Want accurate drafting? Table must account for distance of car ahead of you.
Again, can be a very simple empirically based expression using distance.
Want more accurate drafting? Table must be made for each possible type of car that might be ahead of you and the air turbulance it creates.
Want to race more than two cars at once? A table must be made for every possible combination of cars and distances those cars might be ahead of you for as far as you want to calculate the draft out. Bear in mind each of those cars needs to be running lookups on themselves to figure out how they are driving so you essentially have a dynamic function to lookup a field in a lookup table depending on what another lookup table gives as a result. And that can gets exponentially more complex with each car you add to the race.
Now you are getting in the noise. From a drafting perspective, it doesn't really matter much if you are second in line or last in line. The draft effect is just about the same.
Want to include weather? You must now factor in what differeing (assumed static) weather conditions do to the air disruption of cars around you.
The only dependencies on weather are air density (so altitude, pressure, temperature, and humidity, but then again air density would pretty much be constant) and wind. You are again getting in the noise.
Want to do bump drafting? Add to that a table to lookup associated reduction in drag due to loss of a low pressure area behind your car when another car is close behind you... rinse and repeat all above steps for multiple cars behind you and weather.
Which if I were writing the sim, lead car drag is not affected. Bump drafting involves the trailing car, which has less aerodynamic force to overcome, pushing the lead car. So you want to simulate it? You've now got a contact force you can apply. And I'd say if your sim is worth a salt, it already includes this. Push through the CG and there's no problem. Push significantly away from the centerline and the lead car inherits a yaw moment which can lead to instability. But wait, GTs physics have always been a bit deficient here so I guess they are one of the few games that can't do this...
That doesn't even account for what happens when a car is in front of you AND behind you... lookup table for every combination possible there as well...
Noise level...
At some point, if you want to be detailed enough, it will ultimately be better t use a very detailed model and a fluid motion simulation and then it all comes down to how accurate your fluid motion simulator is
As a practicing engineer, doing both analysis in the design phase as well as in-service issues, which can include failure investigations, I have to ask myself what am I after? What is important? How much fidelity do I need? Each problem is different, but typically cheap linear analysis can get a suitable answer in most situations. If I have to roll up my sleeves a bit, I have to decide what nonlinear phenomena to introduce to better correlate with Mother Nature. As a structural engineer/dynamicist, is it material nonlinear, contact modeling, geometric nonlinear displacements, transient rather than static solutions, strain rate effects? Each of those cost time, and each is a diminishing return.
I can tell you that some very complicated simulations are done in my field of work for little computational cost. Why? Because simplifying assumptions are made based on background work so that you don't have to solve fully coupled fluid-structural-heat transfer problems in a transient nature. Folks running trajectory analysis, which would be akin to the aerodynamic portion of a racing simulation, use data from CFD as inputs; they don't run CFD in-line in their sims. Folks building contact models for complex mechanical systems utilize simple models that solve faster than real-time rather than bringing in full 3D finite element models to get stiffness data. Thermal engineers really simplify their models unless they are doing radiation. Even things like convection, which is inherently a fluids problem, is simplified quite nicely with simple analysis involving convection coefficients (like, from a table!). Dynamicists may work with "stick" models with just a couple of degrees of freedom rather than the large models the stress guys make. Most of your modal mass is in a few modes so you don't need to know that that little piece of hardware rattles at 1500 Hz. It's just not important in the grand scheme.
Detailed sims are great for development. But when it comes to production, simplifications are made based on the knowledge gained during development. Can you get better simulations if you solve the fully integrated system of equations? Sure, as long as those models are validated. Would I expect a game developer to do that? Heck no. We don't do it for billion dollar aerospace hardware. Why not? Because the value added is not worth the additional cost in terms of schedule and people.
In my opinion, what GT needs help with from the aerodynamic perspective is drafting. The draft effect is too great in my opinion. I just shake my head in things like the karting events or the race with Vitz class cars. The AI hits a draft and go into warp speed.
Most of my issues with GT, and this goes back to the original Playstation release, is on the kinematic modeling and in particular contact modeling. That stuff has been pretty poor for the last decade. And that problem is not difficult to tackle more effectively.