Full write access to a database is a totally different thing than reading out the plaintext passwords or getting a leaked dump of the data. Perhaps a mishandled backup.
Which is one of the reasons why these schemes are horrible. But the point remains that banks are afraid of database leaks for other reasons.
Maybe think of it like this: imagine that you have an airgapped system where all the endpoints are running Windows XP (reasoning being something like hardware drivers that were written by defunct companies and can't / won't be upgraded). Is it horrible that such machines are running unsupported, EOL versions of Windows? No question. But if there are other controls in place (like airgapping, like 24/7 physical access control to the endpoints), it might still be possible to provide de-facto effective security.
Maybe think of it like this: imagine that you have an airgapped system where all the endpoints are running Windows XP (reasoning being something like hardware drivers that were written by defunct companies and can't / won't be upgraded). Is it horrible that such machines are running unsupported, EOL versions of Windows? No question. But if there are other controls in place (like airgapping, like 24/7 physical access control to the endpoints), it might still be possible to provide de-facto effective security.