We currently allow analyzed fields to be marked as sortable, essentially because analyzed fields were easier to test than normalized fields in ES (there is a default analyzer, but there isn’t any default normalizer on ES).
Once we have implemented normalizer definitions, we should rework our tests, and then we should consistently throw an exception when trying to define analyzed sortable fields, and tell the user to use normalizers instead.

See org.hibernate.search.v6poc.backend.lucene.document.model.impl.IndexSchemaFieldStringContext#contribute in particular.