Software that has only "bad" parts is also very unpredictable.
(Unless "bad" means something else than "bad", it's hard to keep up with the lingo)
your example is just bad code that unpredictable
My assertion is that software that has only bad parts is way more unpredictable than software that has both good and bad.
For multiple reasons: because "bad" is not necessarily internally consistent. Because it's buggy.
Unless, again, "bad" here means "objectively good quality but I get to call it bad because it's not in the way I like to write code".
Software that has only "bad" parts is also very unpredictable.
(Unless "bad" means something else than "bad", it's hard to keep up with the lingo)