You could take the data and simulate it for different algorithms, and I hope the Steemit Inc team is doing just that.

With a quadratic curve and 4x increase, it could be hellishly abusive, with a whales having the potential to drain out obscene portions of the reward pool. That said, a whale's abuse is easier to detect and police than minnows self voting. On the other hand, the minnow self upvote doesn't cost the reward pool as much.

I have seen a lot of minnows curate actively, so I'm inclined to believe the linear rewards is a net positive.

Come to think of it, the linear rewards would probably be better with a very gentle vote power cost. That'll encourage newbies and engaged curators alike to curate more, and drown out the self-upvote cost. The whole thing about the "so less people have 100% VP" makes no sense to me. I was open minded about it, but I don't see any benefit whatsoever. Engaged curators are basically discouraged to keep curating, while casual self-upvoters who couldn't be bothered to curate others' post gain massively. I know the bots are an issue, but there can be a better solution for that. (PS: For example, voting power tanks after a certain threshold which could only be bots or crazed humans. Currently, the voting power cost actually slows the lower your VP gets! So, let's say, I make 100 votes at the old 40 per day limit, I'd end up with 60% VP or something. The next 100% vote currently would cost 0.3%, which is less than the 0.5% it would cost at 100% VP. I suggest after this threshold, the votes start costing more, and the VP tanks quickly. Of course, the Rshares will need to be decoupled from the VP at this point, so even if a vote consumes 2% VP, the influence given will be the usual 0.5%*VP. This threshold can easily be observed by looking at the data, and should be a weighted percentile of overall activity rather than a fixed number. Just some thoughts, I'm hardly a developer or economist.)

Either way, as a top witness I hope you take this matter very seriously and discuss it with fellow top witnesses and Steemit Inc developers. I hope to see a solution proposed in time for HF20.

One advantage I found is that I am able to give heavy influence to fund projects I created (steemcleaners/cheetah), that I otherwise wasn't able to. This is great, because I don't have to worry for example about paying to run cheetah (she's getting seriously expensive).

You have a very useful service, but other self-upvotes are not so valuable - and most are in fact detrimental - to the community. So, you should count this as a big problem even if you find it personally advantageous.