I could be glib and point out the 'cash' in the title. But I don't have to. Back before the tx volume got big enough to be chronically delayed to the 'next' block, zero conf txs were routine.
There is really no arguing this point. They worked. To the point that freeking _payment_processors_ accepted zero conf txs.
Here's a website that tracks 0-conf double-spend attempts on BCH:
https://doublespend.cash/It is very rare for the transaction broadcast second to be confirmed in a block, even when it offers a higher fee.
And 0-conf can be improved further, for example with double-spend relaying so that merchants can quickly detect fraud attempts:
https://www.youtube.com/watch?v=OAt6B7b4GeMOr by more finely-grained proof-of-work to discourage miners from facilitating fraud:
https://www.youtube.com/watch?v=yXFuNkaYcPQ