I think the best we can do is have an expert familiar with the task and codebase evaluate contributions. Trouble is, this can be corrupted by all sorts of internal politicking.