I think this must be wrong. If the kernel is free of bugs then it's not going to pass a proof of false no matter what the Lean compiler gets up to.