I've always wondered that by gaining the ability to simulate light (or basically anything) in a more accurate manner, we reduce artists' ability to control the final result.

Back in 2D times, an artist could just set a pixel to a given color, and it would be that color, in early 3d, they had the ability to paint in shadows/highlights or control the texture colors directly.

With the physically based workflow, they had to author roughness/metalness and albedo maps, and were hoping for the best that the final result looked good in the game.

Nowadays with simulated light bounces, changing anyting in the scene could affect everything else, making controlling the final result that much harder.

Yes it does look better automatically, but at the cost of adding difficult of channeling the artists' vision.