For me, the closest language currently is F#.
The open-source ecosystem is not as massive as Go's or the JVM's, but it's not niche either. F# runs on .NET and works with all .NET packages (C#, F#, ...). If the .NET ecosystem can work out for you, I recommend taking a closer look at F#.
F# allows for simple code, which is "functional" by default, but you're still free to write imperative, "side-effectful" code, too.
I find this pragmatic approach works extremely well in practice. Most of my code ends up in a functional style. However, once projects grow more complex, I might need to place a mutable counter or a logging call in an otherwise pure function. Sometimes, I run into cases where the most straightforward and easy to reason about solution is imperative.
If I were confined to what is often described as a "pure functional" approach, I'd have to refactor, so that these side-effects would be fully represented in the function signature.
F# ticks the enums/ADTs/pattern box but also has its own interesting features like computation expressions [0].
I would describe them as a language extension mechanism that provides a common grammar (let!, do!, ...) that extension writers can target.
Because of this, the language doesn't have await, async or any other operators for working with async (like C# or TS). There's an async {} (and task {}) computation expression which is implemented using open library methods. But nothing is preventing you from rolling your own, or extending the language with other computation expressions.
In practice, async code looks like this:
let fetchAndDownload url =
async {
let! data = downloadData url // similar to C#: var data = await downloadData(url);
let processedData = processData data
return processedData
}
I often use taskResult{}/asyncResult{}[1] which combine the above with unwrapping Result(Ok or Error).
Metaprogramming is somewhat limited in comparison to Scala or Haskell; but still possible using various mechanisms. I find that this isn't a big issue in my work.
IDE-wise, JetBrains Rider is a breeze to work with and it has native F# support. There is also Visual Studio and VS Code with Ionide, which are better in some areas.
You can run F# in Jypiter via .NET Interactive Notebooks (now called "Polyglot Notebooks" [2]). I haven't seen this mentioned often, but this is very practical. I have a combination of Notebooks for one-off data tasks which I run from VS Code. These notebooks can even reuse code from my regular F# projects' code base.
Over the past years, this has almost eliminated my usage of Python notebooks except for ML work.
[0]: https://learn.microsoft.com/en-us/dotnet/fsharp/language-ref...
[1]: https://github.com/demystifyfp/FsToolkit.ErrorHandling?tab=r...
[2]: https://marketplace.visualstudio.com/items?itemName=ms-dotne...