How about a karma/bittorrent system for researchers to encourage replication? In such a system, the "application fee" for submitting your experiment to a journal would be that you must also replicate someone else's experiment that was also submitted to the journal. This system could scale arbitrarily.