Today, using XPATH with external entities on a Datagrid / Gallery is very efficient. The runtime server knows exactly what attributes to return and can fulfill the request in a single call to the ODATA API server. However, with a Microflow, this same operation requires a lot more traffic to the API server. Unlike with internal entities (which populate the cache with association data), external entities do not. This makes intuitive sense as a default. It makes the Mendix runtime a very good partner in that it only requests data it truly needs, However it makes microflows quite inferior to xpath for working with external data.
What if instead we had the ability to choose which associations (only one tier deep) that we want to populate with a retrieve action. This would allow the developer to give the runtime server a heads up as to what it data it needs to send in the response from the microflow and make the entire process much closer to feature parity with xpath.