> I'm unconvinced the syntax makes a genuine difference

Syntax not, but datalog (-ish) does. It's a more natural way (imho of course) to query data. Far more complex queries with less mental overload.

EdgeQL - https://www.geldata.com/showcase/edgeql

SQL's shortcomings aren't just syntax and the convention of SCREAMING KEYWORDS. There's the lack of algebraic types (particularly galling given the bizarre three-value boolean algebra created by its strange null-handling), the poor composability (why is creating a reusable predicate filter so hard?), the lack of any coherent module system, etc.

The fact that something as simple as a tree is such a nuisance in a "relational" database is ludicrous.