It would be really cool to see a visualization of the spring mechanics that are behind this motion!
As the instructor talked in the lecture, it is really important as well as challenging to prevent the cloth passing through the body.
Based on my experience, even in some very widely used simulation tools, it is really easy to ge wrong collision detection/reaction when the parameters are poor.
I imagine there would be a lot of wasted computation for the areas where the dress doesn't need to be as springy, such as the tighter fitting portions. Is it possible to have a different density of mass-spring particles on those sections? or is a mesh confined to having the same throughout?