[nhaliday]

There have been some efforts along these lines in computer science and math:

http://theoryofcomputing.org/ (has some nice surveys if you're into theoretical computer science)

http://discreteanalysisjournal.com/ (the arXiv overlay model is interesting)

This is a bit apples to oranges—journals are not as central in CS as in biology or other older fields, and the norms about authorship and preprints tend to be more relaxed—but hopefully the trend will spread to more old-world sciences over time.