Would have to be tested in court.

Certainly, the fact that competing encyclopedias exist, and have for hundreds of years, with > 99% identical entry names (but of course, substantively different content), and that predicated not on any given invention or IP but rather the common use English language, would, I think, make the judges rather reluctant to rule differently even should there be 100% match in entries.

Two distinctions come to mind: the encyclopedia text doesn't have a "functional purpose" in the same way as the implementation of an API does, and thus there isn't a market of users who have pre-existing skills with encyclopedia entries that they could put to use if the entries were copied to another platform. In my non-lawyerly reading of the first bit of the decision it seemed they leaned on those aspects quite a bit.

Probably not? The crux of the opinion seems to grant fair use because it enabled a "new and transformative use," which is a box that a different line of encyclopedias doesn't seem to check.

Yeah I was giving an analogy similar to what Google is doing, but I am not sure of the legality of the exact case

Yes.