Hi network-iron,
Let's start with the detailed Model.
Here is a little example of how I would handle the detail shots.
Screenshot 2024-05-18 at 11.29.16 AM.jpg
From that Model, you can render the wave "texture" for the shots with less detail required. If not differentiable, I prefer Texture, Bump/Displacement, and geometry for the very detailed shots only.
As the saying goes, get your money (time investment) on the screen; if it is not visible, don't bother with details. 😉
Of course, a lot is possible if you use Multi Instances or RS Proxies.
I have not seen the storyboard, which would help develop a strategy. Developing an option for all thinkable cases would require more time and may produce too much that will never be used. I guess that is you idea already.
Once the audience has seen the details, then – for the not-so-detailed shots, a little resemblance is often enough to "sell" the shots with less detailed attention. Once a detail is understood, the audience often relies on memory.
Cheers