Some may claim that this is not an objective criteria, and they would have a viable argument. However, this has been the tradition for quite some time now, and it seems to work. The fact of the matter is that the referees have no way of checking the results. Referees do not run the experiments described in the paper to make sure the results are repeatable. That is up to other scientists after the paper has been published. Therefore, there is an element of trust involved. If an author is well respected then that author's results are trusted more than others, be it a new author or an author with a bad reputation.
My understanding of peer review was less about repeating the experiment and more about looking at flaws in the conclusions, overlooked variables, ways to build on the materials.
Obviously not all experiments can be repeated. For example, the E. Coli/Citrate experiment ran for 20 years. No one is going to repeat the 20 year experiment PRE-publication to verify. That would be insane.
To some degree the "results" are going to have to be taken on faith until someone has the time to review the data.
The real goal is to point out: "Hey, your results are fine, but you overlooked the possibility that the cause isn't X but in fact Y which you didn't control for".
Given that Creationists tend to not control for ANY variables, they frequently fail to get published in peer review journals.
™ Version 4.2