I remember reading about the first versions of Eiffel on .Net. That gave the impression that unless you put a lot of effort into your language, it will probably be just a variant of C#/VB. The first versions of Eiffel for .Net lacked multi-inheritance and I think DBC was crippled somehow too. :-/
Your main point is a little questionable - sure, there are languages very similar to C# (Nemerle, J#, etc.), but then you've got F#, P# (Prolog), IronScheme, IronPython, IronRuby... I mean, the existence of the DLR itself kinda puts holes in your argument (for some of the above languages).
Claiming "yeah, but the F# team did everything" is somewhat true (as they did for the first release), but it's not a meaningful complaint. That's going to be the case for virtually any new paradigmatic language being put onto a general-purpose (i.e., procedural, object-oriented) programming framework. The Clojure team had to do everything to make it work on the JVM, too. And in current releases of .NET, more and more of that is being pulled into .NET core and being leveraged by C#.
It's not inherent to the run-time and libraries. IMHO.
I wish Java had been designed with the same multi-language philosophy that .NET has had from the get go. Binaries produced in any language in .NET don't give any clues to what the language the binary was implemented in. For example, if your writing something using VB.NET and working with a library that was implemented in C# there really isn't any evidence that this is the case and the signatures look the same.