You have to remember that a conventional osciloscope is far too slow to capture the sort of glitches which are causing the timing problems which you are seeing,and because they are not repetative even if you use a very high bandwidth one it would have to be a storage scope for you be have any hope at all of seeing them on the screen.I am used to running logic simulations with a timing resolution of 1 ps and its only when you see the circuit running on the bench after a week of looking at 1 ps resolution simulations that you begin to realise just how short a time period even 1 ns really is.