I'd agree that MySQL and others will start to integrate other languages into the RDBMS engines, but I don't think SQL will die. For what it can be used for, manipulating large sets of data, it doesn't appear that any of the OOP or procedural languages to as clean a job. Personally I like T-SQL for set based things, hate it for stringmanipulation, complex business logic, etc.

Somethings are better suited for the db, some are not. The CLR in the database is not a great idea. I understand the function writing, and that could really help in analysis or string type functions, but I'm not sold that writing general ad-hoc stored procedures is a good idea. The database is a limited resource. Better to move these things out to other boxes that aren't limited.