When discussing the algorithm for rendering these three opaque triangles, the layering of different colors is not represented in the target buffer as we only include the color of the object at closest depth. Is there a situation in which we would need to keep this "layered" information? Or is re-rending when the closest triangle is gone inexpensive enough that modern GPUs don't need to cache the overlap information?

Intuitively, I feel like we wouldn't need to keep any of the layered information as a GPU should be powerful enough to re-render this quickly if the top triangle is removed.