My suggestion is that we need a small team dedicated to looking at the economics of Steem, with the ability to stress-test ideas on a testnet and be entrusted to write some reports of all successes and failures.
The aim is to reduce the symmetric design of upvotes so that bilateral relationships are less predictable and hence less profitable. One example already exists on Steem: the curation algorithm and the reverse auction period. That island of activity is unpredictable yet perfectly algorithmic.
I agree that making alternate rule-sets available in test networks is needed - although history has shown that Steem's complexity can result in these test scenarios having difficulty generating realistic data, but it's still better than nothing.
I agree that effecting the way that voting benefits people such that dysfunctional voting is reduced would be a great thing - however, despite a huge amount of communication and thinking on this issue over the years, I haven't really seen any examples of how this can be done reliably and in a balanced way. In the absence of any workable solutions, I look instead at what is workable and hence I look at preventing the function of bidbots.
Curation, as interesting an idea as it is - doesn't actually stop the bidbot corruption of proof of brain because, for one, the curation rewards just go to the upvoter - which might be a bidbot - which still helps the bidbot operator to offer lower priced votes to customers. Higher curation rewards don't necessarily hurt bidbot operators.