That's just how F# team does things. There's no "single direction and decision" behind this - much less people work on F#, it's a much more community-driven language.
Personally, I see nothing wrong with it, because what matters most is how practical the language is when applied to solving a particular task, how well it integrates with the rest of the ecosystem, etc. On this front, I think F# does better than Scala or Clojure, or even Kotlin.
What it does need, I think, is better marketing around data science and machine learning domains - F# is a language which very well fits into this use case, and already has small but nice set of libraries like Diffsharp, and is being used by research labs in pharma sector (albeit on top of Fable, which I'm not a fan of).
You also need to remember that many C# features stabilize very late into release cycle. On one hand, posting about development early and actively brings more feedback, which is good, but it also may give a wrong impression, especially if a casual reader misses "this is early/experimental" disclaimers.
