Regarding the testing of multiple listings...
Initially, when variations on multiple Sitelinks started appearing, I thought it was great that Google considered that a particular client site was so dominant in the field that it deserved the extra exposure. Since Panda, though, maybe even earlier... I've also come to see that this is not always a great thing... that Google is also looking for overlapping material and sites that have too much of it, and it was providing the extra exposure in order to cull things down. I've felt that Google's been using various search refinements to calibrate both its own interface and also to compare user intent and engagement with all other pages it returns for similar queries.
The tough part for Google, I think, is that if people spend a huge amount of time on Amazon, eg, comparing similar product pages (and I know I sometimes do when comparing models, checking reviews, etc)... then search engineers might naturally think that Google should be returning pages in a number proportional to this time. It does make a degree of sense, and I don't think that Google decided to roll out this particular brand authority change lightly, even as a test.
To make a social/political analogy, one which I think goes deeper than just coincidental matching of numbers... in the US, we have the Senate (with two senators from each state), and then also a House of Representatives, where the number of representatives is more or less proportional to population. Fairness turns out to be a complex problem, though, having to do, eg, with the greater relative importance of small towns in sparsely populated states, along with many other issues.
Host crowding is like the senators... two listings per domain, more if it's a navigational query. Brand Authority is analogous to the size and importance of large states.
Please... I don't want to turn this into a political discussion, so lets discuss the model, not the government. ;)