I don’t know why the author is surprised that Postgres offers stronger guarantees than serializability in practice. Serializability per se allows anomalies that would be disastrous in practice: http://dbmsmusings.blogspot.com/2019/06/correctness-anomalie....