Could sampling GGX VNDF sampling,
produce a normal that (based on the view direction) the reflected ray points below the surface geometry normal (i.e cosine of outgoing ray < 0) ?
In other words, my question is, does sampling VNDF take into account the cosine incidence term that appears in the rendering equation? or would it end up sampled rays that goes below the local hemisphere? Specifically looking at Figure 4 (of the first paper) for theta-i = 70, the reflected rays also point below the horizon. Which sugests using the generated normal can produce outgoing rays below horiozn?