Function ensure_unique_pk_index_for 

pub fn ensure_unique_pk_index_for(
    conn: &Connection,
    table: &str,
    column_names: &[&str],
    primary_key: &[usize],
    pk_is_rowid_alias: bool,
    without_rowid: bool,
) -> Result<(), ReplicaError>

Ensure table has a UNIQUE index covering exactly the PK columns, which row-identity point lookups (the engine’s TableSource, the follower’s apply keys) require. Idempotent. column_names is the table’s columns in cid order; primary_key holds indexes into it.

A rowid-alias INTEGER PRIMARY KEY is skipped: the rowid already enforces uniqueness and SQLite point-looks-it-up natively (SEARCH … USING INTEGER PRIMARY KEY (rowid=?)), so a synthetic index would be pure write amplification with no plan benefit. (Databases created by older versions may still carry a now-unused rindle_replica_pk_* index; it is harmless and left in place.)