Nice to hear that I could convince you (maybe partly). I might be wrong, but I really think it should be as you wrote yourself now: diminishing returns (implemented in a correct way) would affect high frequency spammers more seriously than people like you.
You may also think about the advantages of a sigmoid reward curve: It makes self-voting on an 'empty' post less attractive as it is rather flat at the beginning. But it also prevents extremely high rewards (other than n^2 did), because in the end it is getting flat again ...