Keep us posted with your progress on this Royo - I'd love to hear how you tackle it and what solution you end up with. I've got a ~70 project solution which I need to build so, I have a similar problem to yours.
Just to give my $0.02, I just started playing with a build/automated testing for here at work (they'll never adopt such procedure of course)...and felt way strong against using nAnt for the reasons you mentioned. But I gave it a try and it was a real breeze. Within 1/2 a day I have the project building automatically (source/debug/stage) with nUnit test cases running against the builds. Of course this was for a single project :)
However, something I didn't try doing was using nAntContrib's <slighshot> task to convert a .sln to a nant build file. If <slingshot> works as straightforward as all the other nant tasks, I'd highly recommend it.
Are you using the Visual Studio database project to maintain your database schema scripts? If not, maybe thats something to look into. Otherwise, make a habit of always saving database changes as change scripts. Keep them centrally located, and write a little utility (or .cmd file using "for /F ..." ) that iterates over all the change scripts and executes them with osql (or the equivalent for your db if not MSSQL).
I've often thought about, but never decided - for each "version" (schema change) of the database, should I just maintain the change scripts, or should I also reverse engineer the create scripts for the entire DB? Hmmm.... The answer is somewhere in between I think.
Sure would be handy if there were an indicator, down here in this comment area somewhere, as to whether or not HTML is supported in comments ;).