The explanation for COBOL is not an engineering one, but an economics one: “It’s cheaper to train a programmer to use COBOL than it is to rewrite the codebase in <language>.” (Perhaps LLMs might change the economics here.)
"If it ain't broke, don't fix it" also applies here.
There is also no guarantee that whatever new language you port the COBOL code to won't also be seen as obsolete in a few years. Software developers are a fickle bunch.
