>4th) make separate mtdll's out of those objects, so that each object can continuously, independently function, and provide the other objects with data.

Interesting suggestion; it accomplishes something similar to the suggestion that the model exist as a DBF where multiple EXEs, even from different computers, can be adding the new information to the DBF giving us a distributed version of the model.

The big difference worth pointing out between the models is that the one I describe runs deterministicly. It will run the same way every time. I think that might not actually be the case with what you describe, where the objects running independently of each other might lead to them calculating differently depending on the randomness interjected by the operating system code and when each object is alloted CPU cycles.

For the purpose of physics, a determinate model would definitely be more valuable. I think. :-)