I think the issue here is buy in from management. If I work at a Windows-centric shop, it's much easier to sell a different platform if it doesn't require building new servers for it. It's much harder to pitch a project when your requirements include new OSes and servers.
For example, I work at a Windows-centric shop, but if I could make a compelling case to build a tool/product on a different platform, I could get my boss to approve it if it didn't require a new piece of hardware. We have written tools in other languages besides C#, but we stick to languages that are easy to deploy on our current servers.
Windows has C# and all that, and it works nicely for them.
Perhaps this leads to a wider question of who the Haskell audience really is and what problems they think they solve. Lots of languages have pigeon-holes, Ruby has its super dynamism, Python is an excellent scripting environment, C is fast, etc etc. Haskell is... mathsy. Mathsy applications already have Matlab and such.
I think at the end of the day, Haskell will probably remain a language from which other languages learn from, and that's OK, and that doesn't mean it's unimportant (I think even SPJ has said similar).