But if rationalism doesn't at least offer something comparable to other options, many people won't even try.

True. I think the goal here is a bit more complex than "maximize number of self-proclaimed rationalists", though.

Robin's argument in that link seems to be that taking pleasure in disagreement with average beliefs is, all else equal, a bad thing; it's certainly not an argument in favor of updating toward average beliefs.

I was presenting it as an argument favoring updates toward average beliefs over doing so for in-group beliefs, but you're still right that it's really making an unrelated point.

Aumann agreement only strictly applies to ideal rationalists with shared assumptions, but as a rule of thumb one should update toward other agents' beliefs based on the demonstrated rationality of their belief-forming process.

I find such demonstrations quite difficult to identify. Doing so requires both confidence in the correctness of their conclusion and, to a lesser extent, confidence that the beliefs you observe aren't being selected for by other rationalists.