I am an engineer so I know very well that industrial replication is far from easy, but that's a whole different topic here. That's why when manufacturing at scale you implement numerous controls for each step to ensure you remain into "whats known" territory.
Now, when you publish a discovery in a paper, common sense demands that you are sure that you results are not random, and the only 100% reliable way to do that is by replicating your results. There's no way around that. The fact that most people who publish stuff and the system that allows it never ever do that undermines any credibility in said "science".
Hence this is cargo cult. This is typically something that mimics the look of Science by using the scientific method, but it has none of its properties, because the incentives to cheat are obvious and there is nothing in place to prevent them.