Suppose you and I are large institutions. I owe you $10bn today. You owe me 10bn EUR today. I owe you some currencies tomorrow, and you owe me some tomorrow, etc. (I may also owe you a bunch of future corn, you may owe me some shares of Coinbase, etc.).

The practicalities we need to attend to include, at least:

I actually owe you those dollars, and a business decision was made to keep that debt payable today. So I have to convey the dollars to you. Similarly, you need to convey those Euros to me. We can do this by actually transferring them via central bank mechanisms, or I suppose we could have a custodian that operates a blockchain and use that blockchain. I’m unclear how that helps. But keep reading…

There is a risk that one of us will default. If I default before any transactions take place, it’s not a huge deal — I still have my Euros, you still have your dollars, no one is out a huge amount of money, and the lawyers and courts can pick up the pieces later.

But there is also a risk that only one transaction goes through. (Look up Herstatt Bank.) If this happens, then one of us is out $10bn until the law does its thing. That is a big deal. This can be mitigated by atomic multi-currency transactions or maybe by a hypothetical blockchain handling USD and EUR (and Coinbase shares and everything else), but that seems every bit as hard as getting the central banks to use the same database, if not harder. [0]. Or they can use an escrow service like CLS, which works without a blockchain.

[0] Blockchains perform spectacularly poorly in the event of a network partition. Does anyone really think it would be wise for, say, the central banks of the US and Russia to allow currency to move in the same blockchain or, for that matter, on a POW or PoS blockchain?

If we need dozens of specific purpose ledgers between banks, what's wrong with database tables running on ancient mainframes, and how would dozens of different blockchains solve that specific problem?