If you understood why the absence of dead code cannot be detected by testing, and that there is no algorithm that can identify all dead code reliably and in finite time in a dynamic language like Smalltalk, your questions and implications would become pointless, since you then would understand why you never can be completely sure indeed; and that's all there is to it.