My question is why is this not just a demonstration that citation optimization works as expected? They say that resubmission is dominated by a flow from high-impact to low-impact journals. i.e. people submit their paper to the highest impact journal they think the "might" be able to get it into, and then resubmit to lower ones until it gets accepted. This means the scientists are using resubmissions to actively attempt to increase their citation count by getting their paper into the highest impact journal possible. So of course citation rates are higher for resubmitted papers, because resubmitted papers are the ones being subject to this optimization process.
While this is majorly labor-intensive on the reviewer side, it is a decently useful practice in a large field like biology, since it will help insure that "important" results receive a lot of visibility. High-impact journals are nominally high-impact because the community treats them that way. More people read & cite them, so people try to get their stuff in there first and then go to other journals.
I would also note that this study itself is an example of the distorting side-effects this process can have. These guys don't give the above very reasonable explanation for their results, and precisely because of their poor analysis, it is an "interesting puzzle" and gets into a high-impact journal (science in this case). Of course if they'd just done their analysis properly, it would be a completely uninteresting result, and not make it into the high-impact journal. Insane isn't it. That's scientific publication today. Science and nature tend to contain things that are speculative, inconclusive, or plain wrong because knowing the answer is boring.
By the way science and nature are run more like magazines than journals. Most journals don't have an "interesting" cut, just "useful to others, not done before, and done properly". Both science and nature will actually relax the "done properly" part if it is "interesting" enough. This isn't necessarily bad science, it just means that both the proposal of a hypothesis and its falsification are published separately. Nominally this is good for high-profile questions. The problem is that many people don't realize this and often the the proposal of the hypothesis is much "higher-impact" than its falsification!