I completely agree with your take on it; but pls note the entirely different problem Instagram solved vs. Twitter. If you build a real-time streaming application for pure text based objects, your stack will look very different. But collecting, storing, replicating and serving images as your primary social object != same thing.

Hence, even when Twitter 'added' photos, they went with Photobucket so as to not change the nature of their stack. It had taken them 3 years to really get ahead of their adoption curve for 'just' text.