A little off topic.  I have seen similar results (slow .NET access but fast in Query Analyzer) that seemed to be solved with "SET ARITHABORT ON".  Maybe by doing that I am forcing a recompile and it is the same problem.  Of course you need to be careful that the actual procedure doesn't have the potential for an actual error that ARITHABORT will falsely handle.

Hope that helps someone else.  I would welcome any other feedback on a relationship between how SQL Server and .NET are actually handling these differently at a code level.  Maybe there is some other solution that is more appropriate.

Enjoy,

Brian