I think the "selfless/selfish" behaviour terms are not at all times representative and can be misleading so I'd prefer to avoid using them. With regards to the notion of free money, I'm having some reservations using the term as well, especially considering consistently high rewards some users are getting whether or not if PoB is effectively involved. But that'd be an argument of semantics in relation to behavioural expectations, not what reality actually is since we aren't 1-dimensional agents. I'd agree the leak is more flatly distributed across the board, unlike n2 (or any superlinear curve) which congregates such leaks into something more manageable.
Linear could be proven to be a better system than n2 but so far, no such proof exist. Proof that n2 is preferable to linear exist and have never been disproven.
Proof to be a better system in what terms? I guess there are different expectations all around. To me, linear has the shape of a fair consumer product, in that an acquisition and exercise of SP at any level isn't shortchanged by the system in favour of those with higher SP. I'd be pretty pissed off under n2 if I purchase or earn 100 SP, but someone with 10x more SP has 100x influence when it comes to stake-weighted voting, instead of a linear 10x. However, n2 does have its qualities, and on balance, I'd prefer taking only the best parts of it which are useful and hence, settling for something between linear and n2, even if only composed of multiple linear approximations to form something like, maybe, n1.5.
To repeat what I've posted previously about "slight superlinear":-
Firstly, to reduce spam which is evident under linear. Secondly, slight superlinear makes it necessary for all voted content to have a minimum of at least one other peer validation from higher SP users in order for more substantial capital to be distributed, unlike zero validations at the moment. Thirdly, to congregate and amplify the best and worst voting behaviours for community self-regulation, instead of having them distributed flatly and widely like what we're experiencing on the network at the moment. Fourthly, to make vote bidding price discovery less predictable.
I think has commented with some better explanation and insights. So far, that's my take on it..
update: edited a little to clear up some wording..
RE: Time To Wake Up and Fix Steem's Voting Problem