They wouldn't even need to start competing journals - often, there already are alternative journals, or preprints would serve the function of dissemination just fine. The main thing that's needed is that they perform the filtering independently of the publication process.

(Disclosure: I'm part of a project that aims to facilitate that.)