1. Have a place for open publication, like arxiv.org, but with a system of non-anonymous commenting and rating. So if someone puts a paper there, others can comment on it and rate it. Ratings could include several categories, such as clarity, importance, rigor, and so on.
2. There should be a system to rate the reviewers.
3. Out of this would emerge a growing collection of reviewers who are recognized as being very good at reviewing papers.
4. Reviewers would start publishing lists of papers they have recently reviewed, broken down by field.
5. Scientists would subscribe to these lists from the top reviewers in their field.
6. Reviewing could become a profession. A researcher who wants to make sure his paper gets attention could pay one or more high rated reviewers to review it. Note that for reviewers it is important to maintain a reputation as an accurate and honest reviewer, which I think would be enough of a safe guard against authors being able to buy good reviews for bad papers.
If arXiv allowed for signed comments, then famous people could say as little as "This is an interesting paper, and everyone interested in X should read it", and that would do a lot for the reputations of unknown people who are doing good work.
No one is restricted, and everyone knows where to go to avoid having to trawl through the dross themselves.