The game is great alone, but if we install a mod called Real Solar System and another called Realism Overhaul, the force of gravity, engine power atmospheric heights and everything happens to real flyers, then: under these conditions you can simulate a mission there is? What do you think, what other simulators do you recommend?
For all simulators, the core question is: What is it trying to simulate?
Real world space travel has many different motivations, like for instance:
- Military interests
Ultimately, it's the goal that shapes what missions end up looking like. A game-like simulator will have trouble quantifying many of the underlying motivations. Can a computer program easily estimate scientific value of space activity? Geopolitics? What stuff costs and what you can get a budget for? (people often fail spectacularly predicting all of those, even in real life).
With those general considerations out of the way, the game is indeed a physics simulator, simulating:
- orbital flight
- atmospheric flight
- rocket engine operation.
As you have noted, the base game takes quite some liberties with all of those.
If we assume that your proposed mods take care of oddly sized planets and orbits, the orbital part isn't very far off. Kerbal Space Program uses 2-body patched conics to approximate orbits, which is what we care about in most "practical" space-flight anyway. But it still misses out on some n-body effects, like Lagrangian points.
Aerodynamics is a completely different beast, mostly because it's inherently much more complex than orbital mechanics. While the base game's atmospheric model and re-entry heating could certainly be modified to be more realistic, the experience is ultimately not going to hold up to most serious simulations and actual missions.
Rocket engine operation is a mixed bag. While I'm assuming that your mods can take into account the myriad of different propellant combinations, the game fundamentally has to leave out what makes stuff work. Where's the engine development, testing, quality control, assembly, equipment failure, etc.? If all the program does is applying the acceleration vector an engine produces, those same calculations could be done on the back of an envelope. A very very minor part of what managing space hardware actually entails.
Ultimately, we come back to the same question: What are you trying to achieve? Simulators fundamentally select some specific part of reality, and abstracts everything else away. This is often exactly what we desire.
Are you trying to get a feeling of how much hardware is required to get from A to B? KSP can probably help you. Are you deciding the architecture of the next generation orbital launchers? KSP will probably not give you very meaningful results.
The list goes on, and only you knows what parts of reality you want to deal with.
I helped (somewhat) to develop the RSS/RO/RP-1 mod suite (which, respectively, are and do: Real Solar System, which gives you the Sol system, appropriately sized and massed; Realism Overhaul, which gives you historical engines with historical masses, thrust, specific impulse, and rated burn times; and Realistic Progression One, which requires star trackers/radio beacons for vehicle control & provides a career progression & tech tree roughly approximating the historical development of space-related hardware).
The mods that I consider critical for making KSP "simulation worthy" are:
- Real Solar System. You can't even start to be realistic if you're flying between fictional planets.
- Realism Overhaul. Otherwise, you have ridiculous mass fractions and crazy engines which lead to schemes like the famous (but completely unrealistic) "asparagus staging". FAR and RealFuels are required for RO to work.
- FAR (Farram Aerospace Research). FAR overwrites the stock aerodynamics model with one that properly simulates supersonic and hypersonic flow phenomena & drag characteristics.
- RealFuels. Otherwise, you have "oxidizer" and "liquid fuel" instead of LOX and LH2. Models boiloff of cryogenic propellant and feed issues caused by ullage under negative acceleration.
- Principia. This is by far the single most important mod to take KSP from "game" to "simulation". Principia applies a high-speed, very accurate n-body gravitation model to KSP. You can accurately simulate the ISEE-3 mission in KSP. Essential if you hope to do anything resembling real spaceflight. Also models the "lumpiness" of real planetary bodies.
- Kerbalism with ROKerbalism configuration (the best stab at realistic life support requirements yet)
- RealAntennas (highly simplified yet still the most realistic antenna model so far)
- MechJeb & kOS (real spacecraft are nearly fully automated)
What these can't do (a partial list):
- RealFuels doesn't properly model mass distribution within propellant-filled tanks nor the effects of bulkhead positioning.
- RealFuels models tank mass solely as a function of tank volume, not tank area. Ends up being a pretty decent model for most tank shapes (I did a project testing this model against historical data for an upcoming tank materials overhaul) because while skinnier tanks might have the same surface as a wider tank, it carries less propellant mass so doesn't need as much mass to meet structural demands.
- In general, there is no modeling of the vibrational demands of spaceflight more finely than the component (part) level. This is a limitation of the base KSP engine & its modular approach to vehicle design.
- FAR cannot model shockwave compression lift (no droop wing Valkyrie for you!).
- FAR always models the atmosphere as a perfect gas even at high hypersonic speeds.
- Principia cannot model the force of solar radiation or atmospheric drag at altitudes above 140 km.
- Principia cannot model conservation of angular momentum and hence the Джанибеков effect. But this is about to change! egg, the brilliant mathematician behind Principia, has announced as of a few days ago that he has solved this equation & is incorporating it into an upcoming Principia release.
Having tried for years to use KSP as a simulator for developing trajectory optimization algorithms in MechJeb, I'd say there are some severe issues with considering KSP something that e.g. NASA would ever want. Fundamentally KSP is a game and the game physics is based on being a "noodle rocket" or "ragdoll rocket" simulator. That is both its fundamental strength as a game (all the crazy stuff you can build) and its weakness.
The game engine itself is also geared towards game players. The staging engine is not what you'd want in a real simulator, there exists no concept of "booster stage" and "core stage" just a list of actions that gets plowed through. And there exist subtle one-tick bugs on staging where stages disappear and reappear in the engine. The way that game players can rearrange the staging of the rocket during ascent is also absolutely crazy -- NASA doesn't ever fiddle with the staging on the way to orbit, and guidance algorithms shouldn't ever have to deal with the configuration of the vehicle fundamentally changing like that on the way to orbit.
The internals of the game are also somewhat crazy. The way that the "world" is centered on the rocket and travels with it rather than being body-centric, along with the rotation of the world axis that happens around the rocket during atmospheric flight, along with the use of left handed coordinate systems makes everything a lot more difficult (but is really required by the physics engine). And while it may seem like you should be able to do left-handed physics just as easily it changes things like covariant vectors (e.g. "north" vectors become [0, -1, 0] in xzy space) and KSP still uses right-hand screw Keplerian euler angles and I burned around 3 days on a perifocal coordinate change thinking I had xzy problems when I just needed negative signs on all the Keplerian angles.
Of course if you made it a fully realistic and boring rocket simulator then it wouldn't be as attractive as a computer game to the mass of people (e.g. rearranging your staging on the way to orbit was an early feature which was added to the game around KSP 0.21 or something like that because as a game it needed to be more forgiving).
So, if you phrased the question like "would NASA ever build a simulator that looked like highly modded KSP?" the answer is "absolutely not". KSP is built around the PhysX game engine in Unity and around the needs of gamers first and the ragdoll physics is a prime selling point. NASA can assume that everyone using its simulators has a PhD and they intuitively understand that failure to "check yo staging" (Scott Manley) will result in a totally unforgiving failure to go to space.