Thanks for the detailed explanation and the context that you build overall. I agree to most of your points.
Mostly aggregate queries are required for reporting stuff so that query might not be required almost all the time. The point I was making is that SQL DB support aggregate queries and it can execute those aggregates on 100k dataset pretty easily. If I have to do the same stuff in the Dgraph then because of the limitations of groupby we have at the moment, I have to fetch all those nodes to do the aggregation at application level which means fetching 100k nodes just to calculate a groupby stats for a particular schema.