See discussion here. This may be useful to consumers, and it may also help with inference performance. Given the complexity of some of the types in play, esp. around inferring tuple types and similar, any help we can give our consumers is a win!

(These were actually added in a version we support on v8.x, so in principle we could go back and add them, but doing so could be breaking, even if just on accident. Better to do it at the v9 release!)