I think this is a bad way to think about things. Rewarding honest voting means that people have perfect information and act rationally (meaning they understand how to maximize their voting power). The other way to frame things is how systems handle strategic voting. One of the reasons I'm a big fan of STAR is because it both prefers honest voting (you maximize your power her) AND it handles strategic voting well[0]. That is, the difference between VSE_best and VSE_worst is small. Where we see that the gap in Condorcet methods (like RP and Schulze) is quite large (>10%). I think people get caught up in Condorcet methods because they look at RP's VSE being the max. But STAR0-10 only has a 0.5% VSE difference on max but a 4% difference in worse case (not to mention that STAR is substantially simpler and scales better). You can't compare voting systems just by looking at VSE. It is an important factor, but there are many other factors that matter (e.g. handle of strategy, transparency, handling spoilers, scalability, and more mathematical factors).
> Rewarding honest voting means that people have perfect information and act rationally
Democracy itself implies that we expect voters to be informed and rational. Since we know that isn't the case, perhaps the important business of ruling countries should be left up to those who are very well educated and who possess a certain... nobility.
What I like about Condorcet methods is the Condorcet criterion itself. I have yet to find a better definition of a candidate being most preferred by the voters than that the candidate would beat every other candidate in a two-person race. The later-no-harm criterion (that giving a positive rating to a less favored candidate cannot hurt a more favored candidate) is also valuable and STAR doesn't satisfy it.
STAR does seem easy to count though, and without doing much research, likely to elect the Condorcet winner most of the time.
I'm not sure that's a necessary prior. This also isn't a binary condition (informed vs uninformed) but rather a spectrum. One can be reasonably informed (or worse! Misinformed!) and still not vote optimally, but would trend in the direction. Because of this spectrum it is reasonable to also want to handle strategic voting well. Especially because there's nothing better to discourage strategic voting than to make it not very useful (think why would you perform this adversarial attack if it isn't effective?).
While the Condorcet criterion itself is nice it is far from the most important (cardinal systems are going to do a good job here). VSE really is measuring the MSE distance between true voter preference and candidate policy. But there are other criteria that matters. Simplicity is one, as it allows for transparency and voters to understand the system. Ranked systems typically fail the favorite betrayer. This includes RP and Schulze (the most popular Condorcet methods, and what I assume you are specifically referring to). But also consider factors like: expressiveness, spoiling, monotonicity, scaling (do ballots look the same?), and momentum.
It is really easy to get caught up in a single metric, but the difficulty of voting is that there are a lot of metrics that matter and you really have to do a lot of tuning. But I'll tell you that the researchers themselves, including Arrow, typically prefer cardinal systems because of the combination of criteria.
After looking in to it some more, I think one of the strong arguments for STAR is that scoring captures more information about voter preferences than ranking does. The fact that some voters might have a very strong preference for a candidate who isn't quite the Condorcet winner is relevant.
Yes, this is a big feature I like of cardinal systems, particularly STAR and range/score. Although I would take approval voting any day of the year since it captures most things while being simplistic. But I think in an more information based world and where we make policies based on information, a simple voting mechanism where we can capture fine voter preference should allow us to better govern the society (assuming good intentions).
[0] https://electionscience.github.io/vse-sim/vse.html