There might be a way to make manual grading scale. Perhaps set up a cooperative system where students grade others' papers in return for grading of their own papers in another subject matter. By having a larger body of graders, you could have each paper viewed by several people, and it would check that the graders agree on its score (without communicating with each other or knowing how the other evaluated it).
It's just like using TAs, except with casting a much wider net, though a lot more details would need to be worked out.
Remember, you have the advantage that grading a paper is easier than writing one (in compsci terms, the former is in P, while the latter is NP-complete).