I think that conflates two things. The Maps Front End, and the Point of Interest database. Both the classic front end and new front end probably use the same DB. So if the query failed on the new UI, it probably failed on the old one, he just never noticed. Is there any proof that "Sambisa Forest" actually ever worked? Otherwise, it is not a regression.

Google apparently has a "Sambisa Forest Reserve" in it's database, but no "Sambisa Forest"