They do have a SQL interface, and I'm not sure if this is good or bad. It might be that SQL is just a better way of querying, or at least, more familiar. If they continue to use the underlying graph query, then things might perform better. Though, if we're querying via SQL, the complexity returns.

Many have added additional indexing as well, to get around the problems of querying multiple entities and instances. That's where NoSQL type structures struggled. They were much, much better for singleton type queries, but struggled with aggregation. I think some of the indexing now reduces scalability, which to me, means the developers ought to have learned more about RDBMS work in the first place rather than struggling against it.

I do know this might be less about tech and more about people (data professionals) being difficult with rapidly requested changes.