Making line array theory work in the real world inevitably introduces compromises that deviate to a degree from the desired ideal behaviour.
I believe the argument for divergence shading has significant merit. I'm not entirely convinced it's as big an issue as is implied in white papers written...