Does this actually work though? For instance, we may have user data in an API that, thanks to arbitrary rate limits, isn't useful at all when querying in real time.
You mean not being able to access specific user data due to hitting the rate limits of the API? In that case, you can always cache the data (using the cache facilities of the data virtualization layer) and create a hybrid real time/cached workflow that pulls data in real time when possible and from cache for the APIs that have restrictions in place.