I think kdbush should always be faster for points — it uses more than 2x less memory, because when you encode points as bounding boxes, you basically duplicate them. Indexing and search is faster mostly because you operate on less data.
(because sometimes, people have an older, not updated library, for example)