Yeah, I think that’s the idea used in the Ncam/Stype workflows where both systems include a padded FOV value so that you can render with overscan and then distort the render to match the camera.
Regarding the inverse distortion: it’s definitely a big topic and we haven’t gotten into it very thoroughly. I don’t have the exact equations off hand, but we’re just using an approximation for the inverse of the formulas given on the docs page (https://docs.derivative.ca/index.php?title=Lens_Distort_TOP) I’m not sure they are accurate enough for production work, but can be useful for testing and prototyping.