Here's my random musing for the day - If a tree falls in the forest and no one's around, does it make a sound? Or, in verification terms, if a bug exists in the silicon and no one will ever see it, is it a bug? How much time should we spend debugging issues that can never be seen by customers? Alternatively, how likely is it that the situation you thought never would happen actually will happen? Best bet? If it's not a bug, it's a feature! I always like to see these types of issues resolved by adding a VERY STRONGLY WORDED STATEMENT to the architecture spec just to make sure the expected behavior is clear. You could always use assertions to make sure no one tries to sneak something by you during system level testing. Alas, if the customer runs into the issue in real silicon, you're SOL! (Does that translate well? If not, send me a private email at jl at coolverification dot com).