>... and "being a bit of pain" would be true for typed datasets as well no?
True.

Anytime data is represented by an abstraction a change to the underlying data structure will require some additional changes to the abstraction.
When we tied to code directly to the data table, as we sometime did with X-Base, that wasn't always necessary.
I wrote some programs that handle that for me with typed datasets, but I still have to run the programs and rebuild the app.
EF seems a bit more kludgy in that regard, but not prohibitively so.
To VFP programmers, it just seems like a needless pain.
I seen some apps that check the underlying table structure against the abstraction each time they start and adjust for any changes
That seems like a lot of overhead, but that's one way of handling it.