If the library is binary-compatible with that reference implementation (as the typelevel guys say they will be), why would you as a user care which compiler was used to compile it? A lot of typelevel code already uses macros or compiler plugins to generate source (I've even heard that at one point the Shapeless code was generated using Freemarker templates) - but as users we just depend on their released jars and don't have to worry about any of that.
b) I think the part tucked away at the end about a foundation is actually the most important part. Scala's currently owned by one university and also partially by typesafe, and while on the whole they've done a good job there are a lot of stakeholders who aren't really represented there. For a language that so many different organizations are depending on, I think it's time there was something a little more formal and dependable; many other languages have similar foundations.
I agree they can work on whatever they want but this is a bit less benign than writing another library that no one is going to care about. Language forks have never been successful. Please show me one example where a language fork has succeeded and I'll concede. This is a whole lot of drama that is going to turn more people away from Scala which in my opinion is one of the better languages that runs on the JVM.