Exhibit A) SWIFT famously states it doesn't actually move money.
Exhibit B) Hawallah
ie. It's good enough to promise to deliver in due course most of the time. That's equivalent to a transfer, most of the time. The evidence is that it is the basis of many existing settlement networks (SWIFT/hawallah), but also the stock market / individual broker ledger system, the dominant off-chain transfer model of many exchanges for digital assets, crime ("you have X days to deliver Y or Z happens"), etc.
Where it's not good enough, you attempt to store enough to cover eventualities with forward prediction to maintain settlement volumes, offset with promises, utilise third party risk mitigators (insurance/liquidity providers) and/or leverage reputation.
Speaking hypothetically, because the original comment was made predominantly in jest, but in the knowledge that it was actually potentially applicable enough to yield startups, let me humor you. Specifically, in my mind where blockchain might add value is if you wanted to obtain local liquidity at short notice. I understand micro-lending markets are now well developed on Ethereum, but haven't bothered to dig in to the implementation myself. The point is, it would be theoretically possible to build such a system to translate real world promises to on-chain contracts and thus have the ongoing support of globally distributed capital behind providing stability to local micro imbalances of liquidity as a service. This is an existing business model seen in many aspects of the financial system.
Start looking at the world this way, and notice similarities between capitalism, crime, crypto, cold steel and cojones. It's all the same game. Physical or digital, settled or promised, it's about risk and reward, reputation, and "the availability of effective recourse" (ie. trust). The same band-aids are used everywhere.
