[Re-posting. Last post somehow lost all paragraph formatting.]
I was a .NET developer for many years. I launched a production ASP.NET MVC app months before it even reached beta. I wrote a library of HTML helpers which became a part of the popular MVCContrib library. I wrote two MVPs in ASP.NET MVC. Both failed, but not at all because of the technology.
A couple of years ago I moved to Ruby on Rails. The main reason was to flee enterprise development for startup-ville.
Before I explain why I thought I should leave .NET, I want to say I love .NET and C# and think it's the best thing to come out of Redmond. ASP.NET MVC is nice. From what I can tell .NET has passed Java in terms of the language (if not in terms of OSS). And for raw performance you can't do much better than .NET for most applications.
But for startups, .NET has several drawbacks:
1. Ops costs. An Amazon EC2 instance costs 1.5x with Windows. Add SQL server and it jumps to 3x.
2. Developers. It's true, there are a lot more .NET developers to be found in some places than other kinds. But are they the kind of developers suited to a startup? It's not just a question of talent. .NET does not attract as many uber-hackers, but even the really talented devs are more inclined to work in an enterprise context. I know a lot of very talented .NET devs, but I wouldn't bother ask them to join my startup. They make very high rates working for fat corporations, and they are generally not interested in a more risky, albeit existing, option.
3. Bias. You will face bias by investors and influencers who, rightly or wrongly, see .NET as ill suited for startups.
4. Speed of development. Compared to Ruby on Rails, ASP.NET MVC and .NET will take you longer to get your MVP done. It's true that RoR itself makes initial development much faster, but the bigger advantage is in the range and quality of OSS. .NET has a lot more OSS than people think, but with Rails it's almost impossible to think of typical web startup problem that is not already solved with some awesome gem.
Perhaps it might be true that RoR tends to become more cumbersome as a system grows in complexity, especially if you don't have the best devs. But that's a problem that (let's be honest) odds are you'll never face. When you do, you're in a good place to solve it.
It's also true that issue #1 might be too small to matter, especially if you use an OSS database. Indeed all of these issues together do not make a definitive argument. Who would tell the Stack Overflow guys they made a mistake by going with .NET. But all other things equal, I would not choose .NET for a startup.