One can argue that boilerplate doesn't add to complexity (though I don't agree). But no one can argue that it doesn't add to code size. The studies cited in this thread show either linear or superlinear growth of bug count with code size. If those studies are correct, doesn't that rather settle the issue?