> The QA department, then responsible for testing this new "feature" will most likely go off of a script that was sent to them by the developers to verify the old and new behavior, going on some metric that perhaps they don't really understand.

That's wrong right there. If you do this properly the people writing the QA test script work from the same spec the programmers used to implement the feature but they are actually different people and the test gets written first.

Obviously in smaller shops that won't work but anything the size of newsworthy should be set up like that.