It had a couple killer capabilities (to me). I did a lot of work with the Starling engine, and Away3D. And although this conversation is about browsers, a lot of my work was deployed via AIR, which still limps along, but I wouldn't invest in it now. Being able to deploy the same 2D/3D games on any platform and on the web with native GPU access is a killer app - and the best contender now is probably Unity. The ability drive fast vector animations without recourse to SVG or something, all within one dev environment and without many dependencies, provided an incredible workflow from artists to coders. Certain things that were taken for granted in Flash, like the native engine for determining mouse/touch position over layers and layers of different vectors (not bounding boxes) are still incredibly hard to replicate. Whole new platforms had to be written in JS to support these kinds of occlusions and improve rendering performance to a tolerable point. Things we have now like Pixijs would have absolutely destroyed the battery life on a 2010 iPhone.

You're spot on that Apple always wanted native code, and still do, but what small studio has time to write everything three times? The idea of deploying and maintaining totally separate code for a casual game on Android and iOS and the web is a deal-breaker for a 5-person team, and putting all your energy into one walled garden is not what's best for the developer.