I like this approach because it relieves the controllers from much of view rendering responsibility while still leaving the controller the power to alter the view rendering process a great deal.
However, my gut tells me that this approach still hides a non-trivial flaw. I wonder how this approach will cope with swapping out layouts or pages, and actually generating parts of the total view that require increased intricate logic and heavier data requirements.
Whether this approach, a refinement of this approach, or another yet to be discovered, the goal is to maintain the singularity of each controller, adhere to the DRY principle, and provide flexibility in rendering views.