I think this is only a serious concern if the site copies stackexchange. This is not necessary, although it is possible to import the content of stackexchange on a question-by-question basis, without problem.
I was hoping we could do better as a community on this site. The stackexchange model simply wants "question/answer" model, with the focus on elementary questions and clarifications of existing literature. But it is clear that it is easily possible to exceed present published literature, using a model of "submission/refereeing", in that it could serve to actually be like a refereed journal, with the voting process replacing the refereeing in standard journals. This requires careful design to prevent gaming of the system, since if you make something with real world impact, there will be those who will make cliques who upvote each other for increasing impact, and so on. But these problems are in the future, and perhaps are best dealt with in the future.
I am impressed with the visual layout and features of a site like quora--- it is superior to stackexchange as much as stackexchange is superior to usenet. This site, if we have some programmers willing to invest time, can be as superior to stackexchange and quora as they are superior to usenet. There are some ideas that can be implemented without much difficulty that will make the goal of scientific communication more easily achievable. If this happens, the content can be very high quality, and if people are happy with the quality of the text on the site, one could perhaps have an associated actual journal, which can put together the highest quality contributions on a monthly, or weekly basis, depending on the quality of the material.
This can allow a community review of scientific material which is relatively free of traditional academic biases, but it requires a good community, which is able to honestly referee material, so that nonsense doesn't take over, and politics doesn't make the nonsense more prominent than the sense. This is always the danger in such projects, politics is very difficult to arrange to get accuracy.
Anyway, there are folks like Lubos Motl who one can trust to do very vicious review that is usually accurate, also others, so I don't think that inaccurate material can survive very well.
The things that can differentiate this site (if there is a programmer willing to do it--- I am not such a great programmer but I am willing to do it):
1. Good TeX embedding. Stackexchange has a lousy image-based embedding of TeX, that is extremely slow. This should be modified to a quick and dirty TeX embedding that allows hundreds of embeddings without the speed suffering.
2. Good image embedding with local hosting, for diagrams. The remote hosting requirement for png's is ridiculous--- papers and answers sometimes need diagrams. One can also add a native diagram language.
3. One sometimes needs to links movies of simulations, this is not reasonable to use "youtube", as these are not really best represented as mpgs. The simulations in science have a different structure than television footage. These need to be embeddable without effort.
4. I believe the comments need to have no space requirement, they need to be threaded and referenced, like in Quora, without any limitation on space.
5. The voting should not be anonymous, because this produces an incentive for glib voting. The names of those who voted something up, and down, should be prominently displayed. This is so that one can actually see what is going on with the politics.
6. The voting system can be substantially different, in that votes are not sum-tallied, and answers are not collapsed.
I am not sure that the stackexchange model is optimal, the quora model, if you ignore the trivial content, is probably superior in terms of technical site organization. The logos can be completely different, the page organization can be completely different, there is no limit to modifiability.
Perhaps some trusted members can be allowed to do code modifications on a git repository, and this can be tested by the site administrator, and if there is consensus that it is an improvement, the result can be incorporated into the site. But this requires some significant oversight. The model quora uses is not good here, in that any of their developers can make a code "go live" at any time, so that every once in a while, a buggy contribution crashes the whole site.