It's supposed to rely on not-fraud because its incentive structure is not supposed to reward gaming the system (the way that "you get money" incentivizes fraud

It's supposed to rely on not-fraud because its incentive structure is not supposed to reward gaming the system (the way that "you get money" incentivizes fraud in a market system). A certain amount of good faith is presumed because the people who are eventually found retracted en mass will have a rough time getting work in the future, and because they won't be any closer to getting work published that will make them well-known scientists in the future.