Voiceover is currently an extra layer over a visual interface. So, it amplifies the gesture cost of a visual+touch UI.
After writing a NLP interface for a productivity app, I learned that most of the verbal interaction took less time the visual alternative. It didn't matter if you were sighted or blind. Fitt's law meant that navigating a visual+touch UI was too slow.
So, I'm now working on a verbal+touch UI. In that regard, voiceover users are power users.