The highest ROI even for smaller accounts is to find unknown content that figures to be successful once it becomes visible and vote on it.
If there are N low-reward posts the odds of finding one that will be a hit are very low. Is there an effective method for finding posts that "figure to become successful"? If there is why aren't people using that mechanism today? Is the argument that they're not but they would if curation rewards were ~2X what they are today?
Voting mechanically for authors with the best track record doesn't work to generate the highest return because you have to vote earlier and earlier to get ahead of everyone else doing the same thing, which then reduces the curation rewards (the system docks them back to the reward pool for votes before 15 minutes).
Do you have some kind of evidence or analysis for this assertion, or is this speculation?
Well, that is the competitive skill in curation, and there is no cookbook approach to it. There are different methods that can get there from just being a good judge of content popularity to things like machine learning.
A few are, and if you look at the highest ROI accounts by curation that is generally what you will find them doing. A big reason though, is that with 25% curation it is very hard to overcome the natural 4x advantage of just self-voting or delegating to a bid bot. The change from 25% to 50% base curation share is a step to rebalance the system back toward better ROI for curation.
Yes, there have been numerous posts analyzing voting times over the years. I don't have a link handy but the analysis has been done, and it agrees with the game theory (which is itself not mere speculation, it is a theory that is shown to be predictive).