But if you kept moving up and down, you would not consistently have funds to invest. You might not feel you can afford say a new $10k piece of equipment (wild guess on amount) if you may make $5k this month, then only $500 the next 6 months, before making $5k again. But if you knew things were a bit more stable, you could reasonably invest for the future.
That's not saying that longterm planning leads to keeping dictators in place. We don't vote for president every month either. Presumably the 4 year term is not a dictatorship.
There is something to be said for giving change some pacing, and allowing for stability. I think what we're debating is more what kind of timeframe is that? How tight? How loose?
As far as I am aware there is no need to have anywhere near that level of hardware investment to process the blocks needed to be a top 20 witness. Currently only one server can be used to fulfill the role of a witness at a time and therefore there is a practical limit to how much money can be spent on one (rental). It generally makes much more sense to rent hardware than it does to buy it for such purposes and with things as they are (I am open to being corrected) it should be possible for medium to high end standard servers to handle the task at such throughputs. It is possible to rent such hardware at prices that are well within the reach of many if they want to get started in that role.
I previously explained a concept of altering the reward curve for witnesses that would potentially make any challenges here easier to process:
https://steemit.com/steem/@ura-soul/can-we-improve-steem-s-future-by-re-distributing-witness-rewards