If they think it through, and how many potential issues it can have in a mature ecosystem, they'll probably cancel it.
Afaik, they can't cancel it. I presume they chose to optimize away a couple of bytes per transaction and thus they must normalize the balances every 3 years at current 100% annual growth in the money supply.
The reason I was aware of this optimization is I had read about some of Graphene's design decisions back when I challenged the claim of 100,000 TPS and they admitted it was only about 100 - 1000 TPS unless they had very tight control over the witness hardware. Remember I was the (one of the) guy(s) who first exposed that 100,000 TPS lie/hype last year.
What is the bottleneck between 100-1000 and 100k?
(ps, your voting power is almost constantly maxed out... vote some more
)
That not every witness has the same level of ultra expensive hardware and connectivity. That was the case for a more decentralized assumption where witnesses (delegates) could come and go. But Steem appears to be a whale controlled blockchain, thus they can probably make sure the witnesses have a certain consistency of performance so as to handle higher transaction rates. But you are losing persmissionless quality.
Also the other bound is afaics Steem can't be sharded (at least one reason appears to be that voting is a real-time globalized calculation so eventual consistency is not compatible). You've got to funnel everything through one witnesses at a time, round-robin. This is going to cause scaling and reliability problems. Just wait.
Any way, I don't think this is the limiting factor for Steem right now. The limiting factors are more on the economics and voting rewards algorithm, and features of the UI.
I presume smooth or someone will correct me if I have made any incorrect statements.