Blind review to same rubric works reasonably well. Same test with same expected learning goals should remove nearly all issues.