I agree with your questioning of it being capable of preventing bugs, but your second point is quite likely false -- we have developed a bunch of very useful abstractions in "research" languages 50 years ago, only to re-discover them today (no null, algebraic data types, pattern matching, etc).