So, this just cost me a day of fooling around trying to figure out why my image processing effect (which renders some data with a camera) wasn't working correctly.

I get the desire to share buffers between things; but as long as Unity supports multiple cameras it should be assuming we can share the depth buffer between the shadow pass and the regular pass. It's also incredibly frustrating when you come across these types of optimizations, because figuring our what is going on can be very difficult. I'm much prefer the camera system be more explicit about what you want to do, such as being able to Camera.Render(CameraPass.ForwardBase); and know that shows, etc aren't going to be rendered. I often just use cameras to render things into buffers.