Nail, As a programmer (in days past), I want to agree with your hypothesis that today's programming tools are superior to days past, but are still task specific and as a result limiting in practice. I see two solutions to this problem:
First, tools are evolving and becoming more general purpose. Anylogic, for instance, is targeting simulation capabilities in "any logic". While I am not overly familiar with this package, I believe that it's designers are incorporating many useful simulation techniques. The familiar user should be able to switch gears easily.
A web based system such as Forio is also attempting to accommodate multiple simulation engines because interesting Simulations are not limited to stocks and flows.
Second, I believe that a public-private partnership can bridge the gap. People with modeling skills must be matched with people with teaching skills. The partnerships goal would be to create a library of useful simulation teaching tools. Even better the library should evolve as new tools and teaching techniques become reality.
I believe that Barry Richmond described such a library years ago. I am personally interested in this second solution alternative since action can be taken today while we wait for better tools and techniques.