From what I can see from the Palnet posts, the superlinear rewards curve is quite gentle. The proposed rewards curve for Steem is also quite gentle. It is difficult to tell if this curve is an improvement over the linear one. I am very confident it is an improvement over the N^2 curve. I am still concerned that people will use bots to maximise the rewards from the curve. We will have to hope people use their downvotes to prevent that happening.

You mentioned a √N rewards curve. This type of curve should put an end to people using bots to put their content on trending. However, it would probably start a self-voting craze. The large stakeholders would probably split their Steem into thousands of accounts and self-vote thousands of comments to maximise the curve.