I'd say that .NET doesn't favour the uniqueidentifier or guid as an id, but this particular implementation (the ASP.NET SQL Server membership provider) does. I suspect that those who developed the database were working with the assumption that the db usage wasn't to be for high traffic sites, or where heavy reporting was likely to be done.
Perhaps they were trying to avoid any problems with integrating in an existing application, or a future scenario whereby your application had a key for a user. This could be any kind of key for any entity (PK, UserNumber, etc). In the ASP.NET SQL Server implementation, the likelihood of having a collision is very low/approaching zero.
The one drawback that I've learned is that having a clustered index on a guid doesn't scale to large volume databases.
I'm largely in the integer-as-PK camp. They're small, use few bytes, and work very well when your database needs to scale.