Indeed. It seems this wouldn’t be useful for applications where real-time position is needed. You would most likely do the calculation at a later time like in the Miikshi video. It’s a little confusing from the article, but the video actually does a good job of explaining this limitation.
It would make sense that eventually you could do both real-time and after-the-fact calculations depending on whether real-time communications is available. Presumably it will depend on the specific application