I don't get the "no" part: I think you're in perfect agreement with "irrational". There's heaps of COBOL that's very much staying. People are getting hired as junior COBOL developers in 2020.

I mean, COBOL is in the same boat -- there's too much code written in it to outright kill it. They still update the language standard today (2014).

The main migration argument I see is usually along the lines of "50% of all COBOL progammers today will be DEAD by 2025!"

Rather than about technical/financial merit (anyone willing to migrate on those grounds, and believed it was even possible for their infra, will have already done so)