To clarify, none of my issues are with the GraphQL™ layer, I am not using it at all.
The issue is:
v20.11: query runs fast
v21.03: same query runs forever
@amaster507 you have definitely outlined a worst case scenario which would cause… the query to run forever - which it already does. So, which pattern takes longer - a query of a billion that will never finish, or a billion smaller queries that will also never finish?
Ofcourse any happier path will actually return results with the correct pagination+filtering semantics, even among billions of results. Currently, that is only possible if you are ok processing the entire dataset.
Running the entire query without cascade and then again with cascade returns us to semantics present in v20.11, which is the only saving grace here.