You might be imagining two banks coordinating together, now scrap that and imagine ALL banks coordinating, because you don't know which two are going to merge.
Banks have been born at different times (think centuries - or even millenia). So they'd all have to move in lockstep from clay tablets to papyrus to printing press to typewriter before they even decide whether or not they want to bet on computers as a way forward.
To some degree banks already have 'a shared operating system' in the form of clearing houses and central banks. I don't know enough about that stuff. But it's the reason why bank transfers have mostly taken days to complete, rather than seconds, in the past few decades.
Come to think of it, there's a huge incentive not to stay in lockstep with all the other banks if you can offer customers instantaneous transfers when other banks make you wait for days.
This is also why they're often so very expensive. That $25 or whatever isn't being spent on the cost of updating the ledger. It's the cost the bank incurs because of the need to sell assets (loans) before they intended to.
Specifically, the reason that payments are still tied to the nightly cycle is because orders submitted to the central bank aren't executed immediately. At night they suspend submission of new orders and do something called netout or payment compression on the transfer graph, the goal of which is to reduce the amount of money banks need to hold at the central bank. Every payment no matter how small can contribute to the netout, reducing liquidity needs at the CB.
Additionally, there are, like, three big players in the space (at our level, at least). When you’re choosing a core, you’re probably going to be more concerned by questions such as “what percentage of our total revenue will go to this one application” and “what features can we use to save us tons of work and money” and “will this make us more money”. You probably won’t be too concerned with whether converting off the core fifteen years down the line is super difficult or “just” pretty difficult. If you do ask, all three will give you the same answer anyways.
My very very naive take is that they will have had to solve similar problems over time and will have practices and systems that work analogously. I would bet they will probably not use identical software packages or data structures and will have had little reason to structure their internal operations around someone else's standards. This is especially likely to be true of standards that came along well after the companies already solved the problems the standards are aimed at.