In general, FHIR can cover all the use cases of CDA documents. But it can be quite a bit of work to make the transition to FHIR, so most implementations are not doing that right now - they’re sticking with CDA for the established work flows, and adding new workflows based on FHIR in parallel, where the features of FHIR (APIs, Record-level management, better terminology management etc) are more salient.
There’s not really any value in switching from CDA to FHIR unless you are changing what you actually do. FHIR might be a better standard (I think!) but it’s not worth the change unless you benefit from it somehow.
On the other hand, mixing CDA and FHIR works about as well as a person with one leg on backwards. You can get away with it as long as the workflows are quite separate but migrating data from one to the other is expensive.
I think the real issue is that the people in the process could choose to use CDA because they have a limited view of the interactions that they want - based on the ones that they can have - while real opportunities to engage in digital disruption are ignored because CDA doesn’t meet them - the tail often wags the dog here