Usually they have an already existing data infrastructure based on SQL (see Oracle & C.) so the idea of data backup/integrity on the same system is something you usually want. Maintaining another almost unknown data source is usually unfavourable. But it is a trick, because Plone SQL will contain the object serialisation (if Python, with Nick it uses Objection.js which is an ORM).
It is marketing, you say "yes we support SQL" and they're happy to have the data on their SQL server.
My intuition is that the emergent protocol is useful for workflow-heavy projects. I find it hard to say goodbye to Python, but seeing Plone as an API opens it up to use in spaces where it wasn't practical to role out in the past. Plone protocol on Drupal, Plone protocol on Gatsby, Hugo, Astro, Jamstack (Netlify, Vercel).
For some implementations, like Netlify and Vercel, is there a way to do it in a stateless way that doesn't depend on long running processes?
Or maybe I'm just suffering from "having a hammer"