Reproducibility is 10x more expensive, you have to make sure you can replicate all the conditions exactly, and there are no thanks at the end for all that effort. The incentive is to publish instead of finding the truth.