That's a powerful endorsement of COBOL, but a lot of language success is “when did it become common”, “who was sponsoring it or what libraries did it come bundled with”, and path dependence that makes momentarily (due to transient conditions) sensible choices into standards that are mandatory for decades.

I don't think anybody is denying that C++ was (and is) unusually useful, simply by virtue of being the only serious game in town when you need that whole "don't pay for what you don't use" thing, and general performance stemming from that. And devising a better replacement that retains that feature is hard, which is why C++ had so much time to entrench.

But it doesn't mean that we can't do it better these days.