[bitmapper]

I don't really think many people are calling it a panacea. From my experience in the community most people I know who write Haskell write it for productivity reasons (including me, I'd rather be using Idris 2 with algebraic effects though.)

I'm not saying it's somehow magically more productive than other languages, because it's useless to throw around stuff you can't really prove like that, just some people find it to be the case.

I've never found "everyone writing their own DSL" to be problematic, because it's all exposed through the Monad/Applicative/Functor classes, so the way you use them are all very similar. Honestly, the usage of such eDSLs is one of my favourite parts of writing Haskell.

I use Haskell at my work for our main application, and we've been looking at adopting a style guide like https://kowainik.github.io/posts/2019-02-06-style-guide

And about interesting applications in Haskell, I'd consider https://hasura.io/ pretty interesting!

EDIT: And about language extensions, yeah, they can be a problem. What we do is settle on a base set of extensions, and if you want to use another one, you need to provide sufficient justification for it.