Thanks for the explanation. The APIs make a lot more sense in light of it. It seems odd they didn't just use Postgres indexing though. Do you know if they benchmarked it at any point?
For a lot of queries EAV type schemas are really hard to index efficiently. E.g. searching for something like a = ? and b = ? where a and b are dynamic attributes you can't just have a multi-column index when using EAV. So you instead end up with two intermediate query results that then are intersected. If you need even semi-efficient querying EAV usually isn't the answer.
But it is possible to use hybrid data store. Extract important attributes you want to search on efficiently and store them in properly structured and indexed relational database. Other attributes can still be stored EAV style in different schema/database.
