The advice here is not something we intend to follow in general: if there is a problem with our CSPRNG, we should fix that in our CSPRNG rather than doing ad-hoc hashing on top of it.
Additionally, the reference is about Dual EC DRBG, which is not a CSPRNG that we use.
I have left the documents in attic and proposals as-is, since my impression is that those are historical documents that we don't want to try to continue to keep up to date.