I’ve also been getting some strange results (video below), and a general sense of suspicion related to the implementation. It would appear through roughness 0-.998 the specular from the direct light won’t react predictably. It’s possible this is a ‘feature’, but I don’t recall epics documentation mentioning this. It would appear at a glance the the ambient reflections react more correctly through the 0-.998 range.
When >.998 the specular from directional light begins to appear more classically anisotropic (it’s above this point the tangent appears to work correctly, etc). But perhaps I’m misunderstanding. Any insight would be helpful! Thank you