The problem with Apollo is that if you use it for server queries, you are basically forced to use it for local state management as well. This is because server state and local state is usually intermingled, so its not practical to have a Redux store for local state and then use the Apollo cache for server state.

This means that Apollo is essentially incompatible with Redux unless you are willing to duplicate the data in the Apollo cache into your Redux store. Such duplication is a waste of resources, and also defeats the point of using Apollo to begin with, since you are circumventing most of its feature and just using it as a plain GraphQL client.

I think Apollo is trying to do too much, and a much better architecture is not to couple your GraphQL query infrastructure to your state management infrastructure. This can be done by using a more low level GraphQL client and then choosing one of the many amazing state management libraries out there in combination. This way your GraphQL client does not dictate your state management layer.