Thought about this for a while. I also avoided looking up anything on Sapir and his thesis. I am aware he was a hardcore and distinguished linguist though.
I'm getting the sense this is all rather slippery, and there is something paradoxical lurking behind it. I also sense there is a great deal of compression here, a whole books worth. Plus, some of the wording may be leaning heavily on linguistic technical jargon. What is meant by worlds, societies, social reality, or even language? My reason for pointing this out is not to quibble, nor to weasel out of a wrong answer, rather I suspect I'm missing something fundamental. So be it.
My answer is it sounds like a good rule of thumb, yet is not absolutely true. My thinking was, as a counterexample, the many and varied languages found across Europe – most these being dead today. From about the fall of the Roman empire to a ways after the Hundred Years War one could travel some twenty miles, or over to the next village yonder, and not be able to speak the local language. Sometimes that was due to a mere variation in dialect, sometimes it was a really different language.
Thus, one reason why Latin was so important at the time.
If could engage in such travel (assuming you could speak to them all in their own manner), and ask the locals about life, you would get the same answers.
It's slippery though. One might say those two villages are of the same world. How far would you have to travel to claim they are not?
Anyway, wrong answers might better provoke a discussion. So there it is. Someone might want to summarize Sapir's ideas here, or provide a good link supporting all this.