Nicely said... grog to ya for it.
Computer modeling, whether it's the weather, structural engineering, aeronautical engineering, fluid dynamics or whatever, is only as good as the person doing the modeling using the computer as a TOOL.
Therein lies much of the problem that many "lay" people don't appreciate. There are a TON of people using computers tools that DO NOT understand the models or how to apply them. So, they run a model, the computer spits out a number or a pretty picture and blammo...the number or picture gets reported/use/quoted/debunked.
I'll give a little case history from just a couple of weeks ago. I was contacted by a client to do some modeling of something that is VERY dangerous and could have life-ending consequences for both my client and INNOCENT bystanders.
In my reply, I cautioned the client that ANY results we get from our models MUST be thoroughly and carefully tested with real-world, hands-on engineering tests. The modeling results can guide the process, but it's NOT a 'model this, and go into production' type solution. The potential costs are too high.
He absolutely agreed and was already thinking on that page. I like to work with professionals that understand this sort of thing, and only felt compelled to mention it because all too often, the prospective client does NOT appreciate this.
Would you 'ask' a circular saw to build a house? No, of course not. Nor should one entrust the results of a computer model, which are nothing but numerical calculations, to provide a solution to a contextual problem. The computer is only calculating numbers based on the equations it is programmed to use and only using the base data it is given.
Back to the topic at hand....the ONLY conclusion that can/should be drawn by ANY scientist or engineer using modeling is "this model says that cannot be done." He SHOULD understand that the model could be limited (on purpose) or flawed. The discussion should then go into "why might this model suggest that" and the thinking, problem solving HUMAN BEING should evaluate the results.
So, yes, there are many cases where a model says something cannot happen (or happens differently than reality). That's often not the fault of the model itself or theory upon which it is based. Rather, it is either (a) the fault of the USER applying a model in an incorrect context or (b) there is something unknown that the model is not taking into account.
Further, that's how models are improved...just like any other science, it is the FAILURES that teach you more than the successes.
Just some near-random thoughts from a dude who does the following professionally:
Quantum Chemistry Modeling
Molecular Dynamics Modeling
Computational Fluid Dynamics
Reactive Flow (Detonation) Modeling
Author