Yeah everytime someone says "just learn COBOL, you'll make tons of money," it's like,

Bro.

There's a reason no one wants to write new software in these languages anymore, let alone maintain a forty-year-old pile of technical debt.

Been working in COBOL for a decade and this is all true.

I'm lucky. I personally enjoy it. But i can totally see how it's an absolute nightmare for most people.

I've been meaning to learn Fortran in part because because of the whole "big bucks for being willing to maintain old software" thing, but mostly because I'd like to work on the sorts of scientific computing software that was (and still often is) written in Fortran.

Yeah man I'll take plain old php and java any time of day, I can still get enough money from it to pay my lifestyle. And at 5pm I can close my laptop and play vidya with no worries.

COBOL isn't too terrible, it has its gotchas (like sizing variables for inputs (in which you don't need space for the datas headers and will break stuff if you do)) but mostly it's an old language designed to be easy to use

New staff in my workplace first using COBOL (with other build experience) learn it to the point they're productive in a week or two

Been in the industry for 10 years and i deeply disagree with you. I work in COBOL.

Not that migrations don't happen, but in my experience, many, many companies kick that can down the road each year, because migrating huge and critical services is extremely costly, time-consuming and risky. In the short term, just paying people to maintain the dinosaurs is waaaay cheaper.

Also, it's extremely easy to get a job in it ( my company now hires people with no IT background and tries to teach them cobol from scratch ), because even though it's a niche, the demand for it still outweighs the supply of people willing to learn it.

Will it die out eventually? Maybe. I've been hearing about its death for a decade, so i've become skeptical about it in the short-term.

Edit : would also like to point out that it is indeed a fantasy that it pays truckloads of money. Does it happen? Sometimes, but you need to be really good and experienced at it.

COBOL case is bit different. You can't just modernize millions of lines of code that is functionally unique without service disruption - and services that uses COBOL that large often tends to be very sensitive.

The fact that COBOL as a language is both atrocity to either use or read didn't help that either.

Unlike a custom instrument, a dead programming language can be company critical though.