Speaking of automated testing, you should write a followup article about who tests the automated test software? I can't speak for anyone else but we've had the occasional "whoops" occur even there. In the panic that followed to fix the test software so they could meet a deadline, they fixed the original "whoops" and introduced another that was just as difficult to find and fix.
BTW... +10,000,000 for pointing out that even the Actual Execution Plan is NOT a measure of performance. People just don't seem to get that especially when they're comparing two execution plans.