I think it's a serious development tool for
indies. The main problem is that I haven't experimented with enough alternatives to say that XNA is a main contender in today's market...
I don't think MDX is something you should use anymore. It's not up to date and has a lot of holes in its API.
SlimDX seems a much better choice if you want to avoid XNA.
If you look at the huge (and growing) number of games on the "Indie Games" (formerly Community Games) channel of Xbox Live, I'd say it totally is a success. And if you look at games like
Dust:An Elysian Tail (DreamBuildPlay winner of this year), it's safe to say that you can make a quality product with XNA. Now whether you can make good money from using XNA, not as easy to say.
You can if you're serious, have a good pitch and media attention, and possibly if you want to tie yourself to a platform like Xbox Live Arcade.
We have been working on Fez using XNA for a bit over two years now and it hasn't always been a smooth ride, but I'd say that I'm satisfied with most of the API and workflow. I really don't mind the Game framework and the path it kinda forces you to adopt. You can extend it easily and it's just good to abstract some inner workings.
This has been said many times before (in the TV forums), but don't expect XNA to be on the same abstraction level as TV3D; it's not. There's facilities for sprites/models and a cool content pipeline, but you still end up coding much more than you would in a real engine. You have to build your own TV3D on top of XNA, in other words.
But at least you have a good samples library and documentation.
