You don't even need roughness if you fudge the IOR like that. That's my point - not physical reality, at all.
Do you know what dialectrics (i.e., non-metals) in this world have an IOR value close to 1.13? Liquid methane, with an IOR of 1.15! Air is 1.0. Water is 1.33 and most dielectric solids have IORs that fall squarely between 1.4 and 1.6.
The black Shader Ball in your scene, as great as it looks in a biased sort of way, it's made out of liquid methane, right? Because it looks more like black rubber to me (thanks to the roughness applied), which by the way has a real world IOR of 1.52 or approx. the value used for the gray Shader Ball on the right, proving my point, exactly (i.e., blacks appear gray under reasonable lighting conditions).
If the new Physical Material is supposed to be closer to the real world, it should behave like it. Your example of the black Shader Ball solved the "blackness" problem for rough materials, but it's a workaround (and it is a workaround, mind you) that can only be applied to opaque materials.
There is still no solution to the transparency/translucency issue I described in my original post which suffers from a similar IOR issue, even with a hack and even with layered materials. But, I would love if someone can find a workaround for that (i.e., getting transparency to be independent of reflection or at least figuring out a way to tone down and blur those reflections without obliterating clear transparency), because the new material does have some great features that unfortunately are blocked by this reflective "elephant in the room!" And, I am talking a material based workaround that doesn't resort to compositing/post-processing, render masks and render elements, or other non-material based solutions.