Because at $250k/yr they'd have no trouble attracting engineers willing to learn COBOL. There isn't a current "emergency" a la Y2K to spike the demand of experienced COBOL programmers; they have time to hire anyone with a CS background and let them learn the language on-the-job so they can do low-priority legacy maintenance.
> […] and let them learn the language on-the-job so they can do low-priority legacy maintenance.
And that is a problem frequently faced with fresh graduates and young developers (no disrespect to them) whose common reaction to legacy programming languages and platforms is best characterised as «ew, this is gross». Even a heftier salary package does not do the job – they just don't want to learn legacy technologies.
As a personal anecdote, it is the situation I encountered with a few young devs (3-4 years of the industry experience) in my team whom the management had entrusted to learn and support a 4GL platform that they borderline flatly refused to learn and left their employer shortly afterwards to code in React and other NodeJs frameworks. Which is regretful as the 4GL language in question was a pretty modern and good design and could have been learned to get a glimpse into good design and coding practices, the intricacies of the complex transaction processing etc. All of which could have been reused pretty much anywhere outside 4GL. In my view, it was a sorely missed opportunity.
The '08 recession forced me into a job maintaining an old Ada system when I was a young developer. Toward the end of that experience I was put on a project that involved modernizing and updating said system by porting it from SPARC/Solaris to x86/RedHat. I also had to learn to read MATLAB because most of the new modules involved translating the work of scientists into Ada code.
A good learning experience, at least. I've long since moved on to more "modern" technologies.