I would say some type of voting mechanism could be incorporated into the client. All projects could start in tier 1, let's say, and be downvoted into lower tiers as needed. Likelihood to publish could be based on the track record of the project. An unproven project may have to stay in Tier 2 until some type of output has been achieved.

Magnitude/computational power is an interesting idea. For example, should we give more weighting towards projects with the highest magnitude?