Scientific consensus doesn't 'work' in the sense of "when it rejects an idea, that idea is always wrong" any more than it "works" in the sense of "when it accepts an idea, that idea is always right" - what's going on is much better understood in terms of Bayesian evidence, and the apparent asymmetry between accepting and rejecting is actually due to the asymmetry between the evidential status of promoted beliefs versus random possible beliefs.