Message Title

The reason they're actually not pushed to the schema is because when the schema tooling asks the table for the unique-keys, the org.hibernate.mapping.Table#getUniqueKeys() method performs a data sanitize step that is based on the following rules:

its just that there is some sanitation logic for unique-key definitions that map to primary-keys. More specifically:

//  Rule 1
//  If there are multiple unique keys contained in the uniqueKeys Map, we need to deduplicate
//  any sharing the same columns as other defined unique keys; this is needed for the annotation
//  processor since it creates unique constraints automagically for the user
//
//  Rule 2
//  Remove any unique keys that share the same columns as the primary key; again, this is
//  needed for the annotation processor to handle @Id @OneToOne cases.  In such cases the
//  unique key is unnecessary because a primary key is already unique by definition.  We handle
//  this case specifically because some databases fail if you try to apply a unique key to
//  the primary key columns which causes schema export to fail in these cases.

In your specific case, the unique-key qualifies as Rule 2; thus it gets removed. In other words, working as intended.