Mathematically what you wrote in the first paragraph simply doesn't work out. An informal proof: The supply of SBD has to decrease faster than the market price of STEEM, in part because if the price of STEEM did decline, the conversion rate would also decline (you get less value in STEEM for each unit of SBD; the haircut rule). Eventually, if SBD keeps getting converted, this has to result in reaching a state where all of the SBD is used up, and there is no SBD in circulation. (Of course sbdpotato would likely stop first, because SBD reaches $1, but that's okay, it means the amount of liquid STEEM added to circulation is less than it would be in the limit.)
In reality I'm pretty sure that there are multiple factors influencing the STEEM price and it isn't guaranteed to decline just because SBD is converted, as loosely assumed in the above example. It is absolutely possible for the price of STEEM to go up even as SBD is being converted, which would make the conversions even more effective. But even in the worst case, it does have to work.
RE: No-Brainer SBD Potato Top Up Proposal