It feels like the problem is cultural. If the culture doesn't allow enough time for confidence in deliverables, then what proof would you have? Often there are a million items that can go wrong (although all tests pass). So probabilistically something is wrong, but there's no solid item to latch on to.
