Having worked on some major financial web sites (globally), including password code, I can say a few things that may be relevant.

The thing is, you never get a sense of how bad legacy code can be at restricting options in reforming sanity until you have worked on such sites.

It took me about 5 months to restore sanity to one codebase with a bunch of problems regarding encryption and passwords. Fortunately security was a priority, and not just security checkboxes in PCI requirements but real security. But it wasn't cheap and it wasn't easy, and we ran into a lot of unpleasant surprises along the way.

Looking at this the chance is that you have tons of legacy code, and these fit together in not very nice ways. People are afraid to change things because of PCI requirements, security scan results, etc. And the cost of fixing things my be very high. In these cases, I can imagine a "don't rock the boat" mentality developing and a large part of security-critical code becoming effectively untouchable.