Computer Security, The Next 50 Years
There are several "provably secure" computer systems. As in you can demonstrate they fulfil certain mathematical constraints and those constraints are absolute. Then you have to write the code and prove the code, then you have to hope the prover is correct and the hardwareis correct. Nothing is 100%.
As to the randomisation stuff - yes I've got examples, and we've hit the same thing in Linux with randomisation. You get cases where memory scribbles cause a problem only if the layout happens to be a specific variant (especially with stack randomisation). From "either it dies or it works" you get "1 in 10,000 times xyz app blows up". That does make debugging much much harder. Of course a good reply to that is "so improve the debugging tools".