My suspicion is that the "killer app" is an artifact of the super-balkanized era of computing, the era when every vendor's machines ran a different bespoke operating system. That was true in the mainframe era, and it was true
in spades in the early PC era, when the IBM PC and the Mac and the Apple II and the Amiga and the C64 and several dozen vendor-specific flavors of Unix all coexisted, all with different OSes and radically different hardware.
Just from the evidence, it seems clear that porting software is still hard, but nowhere near as hard as it used to be. There's been some convergence in hardware and operating systems. The rate of turnover in these things has slowed way down. But perhaps the biggest reason is that our hardware and the abstractions built on top of it are now so powerful that large classes of apps -- the ones that don't push the limits of the hardware -- have become trivial to port. The term "killer app" was invented around the time of Visicalc; Visicalc was written in highly tuned assembly language so that it would fit in 20K of RAM (no, that is not a typo!):
http://www.frankston.com/public/?name=ImplementingVisiCalc
Nowadays you can implement Visicalc in Javascript and run it on any platform with a customer base larger than five, while routinely creating temporary variables that are ten times the size of Visicalc's entire codebase.
So the modern strategy is to test your software on one platform, and then if it takes off port it to the others. So why don't more people start on Android and port to iOS, rather than vice versa? Maybe it's the vaunted fragmentation problem, but I'd guess that it's mostly because iOS is where the money is. There's an app distribution model and a store. That makes it easier to iterate there. Release the app on iOS and see if it floats or sinks. Then spend your Android-porting energy -- figuring out an ad-based model, testing on many different devices, marketing to all those Android users who don't know what "Android" means -- on the winners.
There are other potential sources of killer-app lock-in. Hardware is one. People who want hardware keyboards can't choose iOS. But that market doesn't seem to have enough size for its killer apps to make news -- particularly because only a subset of Android phones have keyboards, and those keyboards are all different.
There's one very obvious Android killer app: Verizon. Not exclusive for much longer, though, if the rumor mill is accurate.