I think something went wrong with the PR, it shows 4167 files changed...
Yeah, the socket was removed intentionally.
The reason is that many assumptions of the new Fresnel handling, new multiscattering approximation etc. rely on specular and transmission using the…
For zero normal inputs, the most straightforward fix would be to just return N as-is and skip the process. The entire algorithm is undefined for that case.
I'd go a step further and argue…
Generally this is good to have, but I think we should a) support the full range of Microfacet distributions and b) have an enum for switching between F82 Fresnel and the proper conductive Fresnel…
I've found one case where I don't think the behavior is expected: When using a variable Metallic input (e.g. a texture) with a fixed Specular Tint (e.g. 0.5), the code produces two Mix nodes, and…
Seems reasonable - EEVEE isn't doing the diffuse/SSS closure mix that Cycles now does, so the old behavior is probably as good as it gets for now.
Fixed with #114300 / 0b81a163e2560a3f72a512ad00d101af54081868.
Not updating automatically generated conversion setups is accepted as a known limitation for now.