In 06b0475f172155517908c256f579897783c60a95 , for
DBZ-2911 , a "sequence" field was added to the source metadata stanza, containing an opaque list of numbers identifying the record's position in the stream, which can be used by downstream consumers to identify duplicate or misordered messages.
For Postgres sources, the sequence numbers are intended to be (LSN of last commit, current LSN). However, the current LSN field is never updated, so we are actually recording (LSN of last commit, LSN of last commit).
Thus if a given SQL statement generates multiple records, they will have the same sequence field, defeating the purpose. For example: INSERT INTO test VALUES (1, 2), (3, 4);