[wdobbels]

At least in astrophysics, the standard is to let your paper go through peer review first, then post the paper once accepted. In the arXiv comments, you then put "Accepted by <journal>". Besides offering the papers for free, the arXiv is the main "news feed" that astronomers use (much more convenient than having to check all the different journals). Imo this system works quite well. There's indeed papers posted that are not yet peer-reviewed, but these you read with a more skeptical view.

The large number of submissions is indeed a problem for other fields (I think mostly computer science?). I'm not sure what the best solution is for that. A voting system (where you see a mix of new and popular submissions), together with some content filtering, can help the reader. However, this would probably lead to some important papers getting burried.