Given the nature of the breakage (tripping over a non-w3c standard heuristic in one browser with no obvious and visible error resulting), I'd assume no engineer knew the code could trigger problematic behavior for some users until I see evidence of intent to the contrary.