It has always been about two things - getting lucky, and moving fast.
Playfish tried to do a lot of games "right the first time" but weren't set up to quickly fix their mistakes and have ended up performing terribly.
On ratings-focused sites like Kongregate and Newgrounds we might literally only have hours to measure, judge and react to feedback and bugs before our game slips off the new games page or the ratings-damage is too great to recover from. And that's if we're lucky enough to be given a reason why "this game sucks 1/5" heh.
On the rest of the internet we have the issue of games being distributed beyond our control, once a game goes 'wild' it gets to 100s of sites and there's really only a couple of options for pushing out updates - hot load an x megabyte swf to millions, or Mochi's LiveUpdate system which requires their ads (sometimes we're not allowed to use ads).
Sometimes only a handful of games satisfy these requirements a month, out of 100+ a day.
Social games I imagine are much the same .... you have to get a lot of things right first time, and your list of things that were right yesterday aren't necessarily so right today.
Really a lot of things are the same ... the difference between making a product your friends like vs. your town vs. your city vs. your state vs. your country vs. the whole world .... probably far below 1% of 1% of 'stuff' makes it to the top.