Agreed. Many noSQL-boom-era databases eventually bolted on a SQL-esque layer, but that was also because they were mostly also all targeting "enterprise database" use cases and customers who both expected that and whose use cases largely fit with it. So, there was a lot of pressure to conform to norms when the advantage of not doing so wasn't immediately self-evident.

We have a database [1] and query language [2] that's tailored to storing & querying trace/telemetry data produced by different layers and components of cyber-physical systems for systems engineers to analyze, verify, and validate what a complex system is doing. It's not quite a traditional relational problem. It's not quite a traditional time series problem. It's not quite a traditional graph problem.

Addressing the way that systems engineers think about their domain in an effective way required coming up with something different. Are there caveats and rough edges? Sure. But, they're a lot less pernicious and onerous than the alternative of trying to leverage a bunch of ill-fitting menageries of different solutions.

Redis is fit-for-purpose. So, it makes sense that its query interface would also express that.

[1] https://docs.auxon.io/modality/

[2] https://docs.auxon.io/speqtr/