The UNION / UNION ALL is a good one (and John's "trick" one about multiple clustered index definitons). Sadly, about 5 years ago I almost had a knock-down, drag-out fight with a DBA who thought UNION and UNION ALL did the exact same thing.

Here are some other ones that can help to seperate the wheat from the chaffe:

- Can they name the 5 transaction isolation levels, and the behavior of each one?
- If you issue a single UPDATE statement that affects 10 rows, does the update trigger fire once or 10 times?
- By default, if an update trigger contains code to update the same row that fired the trigger in the first place, does the update trigger fire again?
- What's the differerence between an identity column and a uniqueidentifer columns, and how do you populate the latter of the two?
- What's the difference between a DELETE for all rows and a TRUNCATE?
- Why can you use an expression alias in an ORDER BY, but not in a GROUP BY or a HAVING?
- What's the lifetime/scope of a common table expression?
- In SQL Server 2008, what's the new command that allows you to synchronize the rows in one table against another? (just to see if they're paying attention to the new language features
- Assuming a table called Orders with a datetime column of OrderDate, if I issue the query select * from Orders where OrderDate between '2009-02-05' and '2009-03-31' , by default will I get all orders placed on the dates of 2/5 and 3/31? (This one catches many people)

Obviously, even a good candidate may not knock all these out of the ballpark - but in general, can help in determining whether you want the candidate working for you or your competition.