Perhaps some more options for content discovery, filtering and such ought come first? Before a major economic shift like 100 Curation 0 Creator reward?

Filter by word count, combinations of tags, and just more that five tags in general. Search granularity isn't fine enough. "Not worth it" on average. (Steemit has market position; it's where everybody new goes first.) Search cost must be far less than value-to-the-reader of content found. Which not the case, and that, besides the boundary conditions, is possibly half of what incentivizes many large and small stake users to farm. (Thereby lowering the prestige of the system.)

Then a "recommended" algorithm with parameters users can tweak may be the first thing to try.

Agree about the linear versus nonlinear rewards, basically. Same argument as you cited.

Overall if doing economic shifts, whatever is chosen must be strategically neutral. If the desire is for it not to be gamed. Meaning that if a user makes a list of alternative events and estimates probability and their strategy is the aggregate of best responses to each of these; but it shouldn't depend on which of them actually happens. And best responses must exist for each. (The sleeping soundly rule.) The strategy should not vary depending on which one the user anticipates will actually happen. (The user should not have to believe anything.) And no strategy should be both pure and dominant.