RE: A Method To Combat Upvote Spam Aimed At Maximizing Curation Rewards
This already exists. Every vote has a dust vote reduction in order to discourage vote spam. Very small votes are worth less and earn less. A small to moderate stakeholder who splits into 10000 0.1% votes per day will give up a lot of their vote power, possibly all of it (on this note, also see third paragraph).
You're talking about the 50 rshares reduction?
Account splitting is another issue, and no the 3 SP isn't a major obstacle because: a) empty accounts can be bought in bulk, and b) if the strategy is significantly profitable then the 3 SP is just a start up cost which would eventually be recouped anyway.
Is there a store of empty accounts somewhere someone is willing to sell in bulk at a large discount? Sure, the 3 STEEM (not SP) is just a start up cost.
Also, many of the dust votes you see are in fact pointless. They're made at zero seconds which means they forfeit any curation rewards even if they earned any. I believe you have inferred some sort of intelligence or strategy behind them when none exists. I don't know why people are doing that, other than ignorance about how the platform works perhaps.
Of course, I'm not talking about those votes that are given at or very near zero.
You should take a look at the user @eforucom. I've been following his rewards on SteemWorld. Maybe you have tools ready at hand that allow you to look at this user's curation rewards and voting behavior from a longer time period. While his curation rewards haven't been anything to write home about very recently, I recall seeing weeks when his curation efficiency was greater than 150% prior to HF 21. Lot's of relatively small automated upvotes. Nowhere near the vote dust threshold, obviously, but quite small anyway. A lot of the curation taking place here is automated. That's not too bad if the authors are picked with some care.
BTW, I added a tiny upvote to your post even though the idea is mostly if not entirely wrong, because it is good to examine such things and raise questions.
I think automated and thoughtless curation, which was one of the issues raised here is not "mostly wrong". That includes things like front running bid botted crap content.
I don't know the exact amount. I remember it being described as a small vote on a small account. Whatever the number, I know it is significant because until the latest hard fork, people with small accounts did encounter the situation where they weren't permitted to vote because the deduction took the vote to zero (which was changed in the fork, unfortunately being part of the new conditions which led to one of the chain-crashing bugs).
EDIT: I looked this up, it is 50 million rshares, not 50 rshares. It corresponds to an account with about 1 SP making the typical 10 votes per day, an account with about 10 SP making 100 votes per day, etc. Massive splitting down to, say, 0.1% votes by a moderate stake account will destroy the vote power, and if you see someone/bot doing that it indicates not a successful strategy but likely a mistake.
Funny you should mention that because I have seen a literal web store where these were being sold. I don't know if this still exists, but someone wanting to milk the system can surely find a source of such accounts by asking around. I occasionally get solicited to buy them, although (note to would be solicitors:) I don't have any interest.
There is in fact, currently, limited demand for this specifically because the system is designed to avoid introducing incentives for splitting (although there are still some), and when developers do find such incentives they try to come up with ways to remove them rather than adding more.
Your post is mostly wrong in part not only because it is unlikely (or at least not well supported) that anything harmful exists, but because the suggested "fix" introduces account-splitting incentives which is contrary to an important Steem design principle.
If you: a) have better support for the idea that there is a problem worth solving other than the observation of some votes existing which are likely automated and have a nonzero return (this is fully expected and by design), and b) have a proposed solution that doesn't introduce/increase incentives for account splitting then you may be on to something. But right now, you aren't.