For one thing there are a lot of people who are not qualified to have an opinion who would like to chime in on the topic and in certain fields (say gravity) they outnumber the real researchers. Some of these people are clearly psychotic but some of them will put up an huge fight, in the courts even, and other people are marginal cases, right on the edge, where you will lose the moral battle if you keep them out.
Another issue is that some senior (and junior) scientists tell it like it is (or rather how they think it is) and don't stop to think about the emotional impacts their words have on people, particularly when shared with the wider community.
These people have valuable things to say, but they make mistakes like the rest of us, and the peer review process defuses much of the tension in this situation, although it too has the problem with tail cases with very high cost.
The crux is figuring out how retain the good aspects of the existing system (well-considered and appropriate peer review, approximate ranking by importance) and getting enough traction to allow it to take off.
I was involved with arXiv.org and we estimated costs somewhere around $5 a paper, however we were not doing any typesetting (mostly we got submissions in TeX) and no peer review. It may sound blasphemous, but it is not clear how much value you really get out of peer review.
One key is that arXiv was running on a very large scale, if you are considering the cost of the programmers, hardware and such, it is just as expensive to run a small "journal" as it is a large one.
The reason I don't work there anymore has to do with the extreme difficulty we had in finding that $5 a paper. After years they've finally been able to get other academic institutions to put some chips in the pot, but it seems to me that it is much easier to find money for a boondoggle than it is for a lean operation.