Scrape SDK Distorted Output? What Exactly is in the Runtime?

Looking to mirror an SDK distorted OpenGL game. I understand it mirroring is currently accomplished in DirectX/Unity by setting a flag and then handled in the Oculus runtime. Seems like it might be possible to scrape what the documentation calls pRenderTargetTexture for the last frame prior to writing the new one? If not seems like being able to send a framebuffer into the API to get the final distorted image would be useful here and probably other cases as well.

I haven't dug through all the Oculus code, but is there a good explanation with order of events anywhere of what and how it takes place in the runtime vs the SDK when using SDK distortion?