I don’t believe that that may vertices is too many. My record is 2.6 mill verts, but then again, I had to layer things over 8 layers so I could have enough CPU to model and I got an extra gig of ram. It also took 8 hours to render one picture. I never finished it, which was good, for I was a third done when I stopped
I know in movies what they do is render the setting, then they render the characters with the floor. That way, they can render the background, which does not change (unless a character casts a shadow on an item. In that case, the item is moved to the action layer) and then overlay the character action layer. This works extremely well since the lighting and camera movements are identical. As well, if the animator wanted to change the action, only the action layer would be changed and then the animation of the background would be reused. Otherwise, the computer would have to render out the background and the character every time a change was wanted.