Google Developers Group
I will risk saying something different then everyone else; Android versions pose no longer a problem.
Based on the large statistical sample of our popular apps I can dismiss what used to be a big issue for Android lack of upgrades. Currently, our statistics show that there are only 2 perfectly well working Android OS versions that we should be developing and testing for:
Android 2.3.x ~ about 60% of the market
Android 4.0.x ~ about 30% of the market
All other versions, especially Android 2.1 and below are less than 1% and could be ignored unless you have enormous financial resources and strive to carter to every last one of your customers. Android 4.1.x is of course emerging.
So where we seemed to lessen the OS version problem the Android device screen size seems to be getting worse as for example with Galaxy Note and Galaxy Note 2, were Samsung decided to change pixel density and take few dozen pixels off the screen just to make things more fun.
This may minimize the problem from the developer's perspective (although I would still like to reach the largest number of prospects possible), but it does not affect the problem from the user's point of view, and many developers have friends (not to mention the developers themselves) with older versions of Android. Also, if a developer's app can no longer be updated because the latest version will not run on the user's device, the user might be less inclined to use other apps from that developer.
I think you miss the point, its not due to android OS version market share but the sheer number of open source libs that have back ported android 4.x OS look and feel features to android 2.x.
github.com search is your friend
my analytics from Flurry seem to support this as well.
Android 2.3.4 has been on the top of most of my lists this past year.