In PostgreSQL each schema type is stored in its own files with its own index, and every record is accessed separately. In ZODB, a single pickle can include many differetnt classes. Pickles load fast, as a single object access. In contrast. accessing objects with multiple schemas in Postgress requires accessing multiple files. That is slower. In practice, to minimize load times, PostgreSQL developers self-limit the complexity of schemas, making applications more brittle.
But Pickling offers more than just fast loading. It offers dynamic schemas. Postgress schemas are relatively static. Pickles can be far more dynamic. In PostgreSQL every schema has its own definition, and files. I believe that you can subclass schemas but that generates a new set of files and indexes. In PrivaCV every person has a custom skill object, with many different parent (Z)Classes. The user chooses the skills, that defines their custom skill class. I have a hundred skills, soon to be a thousand skills. Imagine how many PostgreSQL schemas that would require. Every PrivaCV person skill object is a single object, part of a larger pickle, and loads in a single object access, no matter how many specific skills in inherits from.