For 1, there is a recent proposal [Pre-HFTP] GHC.X.hackage: A tool for easing migrations to new GHC versions that should make it possible for HLS to support new GHC versions much faster. This seems really promising, so I still have a lot of hope that this will mostly be resolved in the near future.
For 2, there has been a proposal for the wiki in particular [RFC] Evolution of wiki.haskell.org, but no clear plan has formed yet, I believe. I think it is difficult to get consensus on what the best course of action is. As for general documentation, I think the Haskell foundation has set something in motion to improve the documentation, but it is still going a bit slow. If you have concrete examples of bad documentation, please report them as mentioned before in this thread (thanks for making that repo @tomjaguarpaw!).
For 3, I consider all your examples to be in the data science domain. That’s a domain that the Haskell ecosystem just isn’t very good at yet. As mentioned in this thread, there have been attempts before to improve this, but it hasn’t been successful yet.