You would think that this setting would be optimized in such a way that after the PrePass it would just finish the render and return the Depth Texture, but that’s not how it works. No, instead it STILL goes through every pass regardless.
So instead of taking 0.12 ms to render it takes 2.43 ms to render… 
That means capturing every frame to a texture and using it in a material to create, for example, cool post processing effects is completely impractical, especially if you need multiple Scene Captures or, in my case, Scene Capture Cube which needs to render the scene 6 times.