I wonder if the "programmer" (and I use this term very loosely) who wrote that sleep-in-an-interrupt code ever tested the code personally, or if it was some other distant responsibility-diluted department of a hundred other lamers who didn't care "because the automated tests all pass". This is a situation where dogfooding, in the original Microsoft sense, would definitely be beneficial as among the developers experiencing this on their own machines, surely one would be tempted to fix it.