There isn't a lot of support for superlinear now that is true. It would work better with cheap downvotes I agree there. Even the original n^2 with no whale experiment probably would have worked better with cheap downvotes. The original white paper talks about downvoters pulling down the top overpaid posts (the famous crab bucket analogy), but in reality it rarely ever happened because the cost of doing so was too high. Address that and even unmodified n^2 might be significantly better.