For example, I've seen time and time again, pages on my site that got hit by Panda rank solidly in the top 5 for 'blue widget' but is nowhere to be found in the top 900 for 'small blue widget' even though the page title tag and h1 tag actually read 'small blue widget'.
I can give you a counter example. I have a small site that was just pootling along without much input from me (I had bigger stuff to worry about, because I was working on a pandalised site), and suddenly when panda 2.3 happened (July), it surged to #2 for a term it was not optimised for (because this keyword never occurred to me when I first made the site).
When I say not optimised, I mean the phrase was not in the title, the url, or any of the headers, and it wasn't in any anchor text in backlinks either (I checked in case someone had kindly built one for me!)
It didn't even occur as a phrase on the page that they chose to rank - though the words that made up that phrase did occur loosely in a 12 word sentence (which G has picked as the snippet for this particular SERPs). [Edited to add: my page is definitely relevant for this search, they haven't ranked it because of some random occurrance of words]
It's remained ranked, and it converts nicely into sales.
I've been racking my brains to work out how it happened, so I can duplicate it, but it's beyond me. Unfortunately this hasn't happened to any of my other pages, so I've no other example to compare it to.
All I can conclude is that G did one of it's experiments - tossed the page up out of nowhere to see how it did, and then ranked it based on user metrics.
I think there's a luck element to Panda, because if you haven't been tossed up as an experiment, G doesn't get to gather enough data to reward you (though of course you could fail whatever test they are setting).