If users find this horribly annoying, we could dedupe results, but I'd prefer not to.

It's possible for the same result to get a different score with different titles. I'd like to be able to mine the data to see how often that happens, and whether it follows any obvious patterns, and what it tells us about how we display results.

A good example right now is "Aimee Osbourne" (daughter of Ozzy Osbourne). There's a redirect from Aimee to Ozzy, but CirrusSearch only displays the title "Ozzy Osbourne" in the results (I think because of the partial match in the One True Title). Google, however, shows the title as "Aimee Osbourne". It's the same page in both cases, but I bet people are at least somewhat more likely to click on the obvious exact title match.

Another example: searching for "corn" returns "maize" (with a redirect from "maize corn"). I used to think "maize" was a kind of corn, not just another name for "corn", so old me might be less likely to click on even "maize corn" than plain "corn".

Mining for other examples and seeing how people rate them would be useful. I'm hoping I can use it to convince us that we should always show exact redirect title matches if we have them, and maybe more generally work on scoring and sorting redirect title matches.

Or it could turn out to not be consistently different, in which case I might drop the matter (though maybe not).

I think this would be really good stuff to take a look at, @TJones !

Moving to later this quarter....just not a huge priority right now.