Even if this is the only effect we see, it still has pretty major consequences for the future. How would one become a senior developer in the future if there are almost no opportunities for junior developers?
The question is what happens when all the graybeards are dead. Are the institutions that are refusing to replace their systems banking on future AI being able to solve all their problems single-handedly? Or are they going to offer a pittance to any one who chooses to learn COBOL? Maybe it will be something in the middle; they will only hire senior engineers with specialties in developing AI/ML models for legacy systems. That is until those seniors become obsolete by their own creations.
Ultimately, you can start a new bank with new technology, move clients money to new technology, have the old bank go belly-up and bankrupt. Sometimes that happens in a new company, sometimes it happens in the same company (i.e. re-writes).
Maybe not so extreme, but I think that's likely what will happen over time. Maintaining old software is a choice, and is sometime more expensive than it's worth (depending on what that software does).
It can certainly be very painful in certain cases (the Great Depression was a similar "reset" for the financial system as a whole - though without code involved), but the market mechanism will accomplish the objective, though likely not without collateral damage.
They will go the way of the typists.