The way this could work out well is if the professor is the long term contributor, and assigns students appropriate tasks. The class would then essentially compete for the best implementation to be submitted upstream.
But this is broadly the same concept as the Google Summer of Code, with the caveat that instead of the students pursuing something they're specifically interested in, these students are trying to learn a skill. Rarely is the result of learning some wonder to behold (to the contrary, most learning is as a result of failure).
Having more students involved in open source concepts and communities is certainly a good thing for open source as a whole, but unlikely to be a panacea without significant cultivation on behalf of this professor.