That is odd that the merge should be faster than the hash, since if I add hash to the join (i.e LEFT HASH JOIN) it executes in half the speed...

Both systems have the exact same data (well the test system is about a week old, but only 1% different?) I am restoring to test now to test the exact same data, but nothing has change in the last week to warrant any performance issue changes.

Grant, I checked the server default options and they are both the same for each server.