If reproducing were a requirement for reviewing, publication cost might be exorbitant (like when CERN tries to publish a paper you have to rebuild the particle accelerator?)