The correlation of EAV schema’s and JSON columns is a great point, and I think the post does a great job of simplifing the problem to show exactly how similar the two approaches can be.
But I can’t help but think EAV presents many different values then JSON as your code base changes. For example if you look at the Zotonic it’s data model is an EAV schema, based on something like RDF triples. It stores the binary of the code terms in a column… which I’m not a fan of at all. But it does have a very nice system of pivot tables which point back to the “Entity” resource table. This seem much more natural to me than either he JSON or embeded binary.