Having worked for a bank, I'm shocked that they managed to build a product that offers no business value but requires actual technical expertise to implement.
Now that I think about it... it's probably an Excel spreadsheet...
I upgraded a Visa Level 1 merchant's credit card system a few years ago. It was an unholy mess of C++ servers, Java servers, large amounts of semi-comprehensible PL/SQL, and Korn shell scripts. Real time authorization was handled by one system, and end-of-day settlement was handled by a very different system, under a different director.
I wouldn't be at all surprised that the "real time" data the feds get is very nearly worthless. It's probably patchy and intermittent, and varies greatly in timeliness. The document says that "open ended" systems (Visa and MasterCard) depend on issuing banks, which is certainly true, but the merchants can also contribute greatly to delays in getting data in. The Level 1 merchant did something called "Assumed Approval" when it lost contact with the payment provider. Since about 96% of all authorization requests got approved, it would just assume that all auth requests got approvals when network problems prevented actually getting approvals. The feds wouldn't find out about those transactions until end-of-day settlement.
Now that I think about it... it's probably an Excel spreadsheet...