Very cool, if true

A colleague of mine spotted a bug yesterday in a piece of test automation, in which a specification of a list of test cases had contained some tiny easy-to-miss lexical error (I forget exactly what, but something along the lines of putting an operator inside rather than outside a critical pair of parens) and as a result two long sublists of test cases had been fed to the Cartesian product operator instead of the concatenation operator.

It was only two lists being combined, fortunately, but the same sort of typo could just as easily have done the same thing to n lists, which would have turned a more or less linear amalgamation into an exponential one.

(Though I suppose the worse it gets, the more likely you are to actually notice when you investigate why the test system seems to be spending forever running indistinguishable variants of the test suite in question. Perhaps the real limiting factor on this question is not 'how egregious a consequence can you imagine for a trivial code error?' but 'how egregious a consequence can you imagine going undetected for some reason?'.)