Can someone shed some light into how the deduplication works. What does this query actually do?

CREATE TEMPORARY VIEW IF NOT EXISTS subjects_distinct
AS
SELECT *
FROM (SELECT *,
             ROW_NUMBER() OVER (PARTITION BY `id` ORDER BY updated desc) AS rownum
      FROM subjects)
WHERE rownum = 1;

I imagine it turns a kafka message stream into a changelog stream. Does this mean it keeps track of duplicate ids and issues retractions when one is encountered? I tried testing this, but didn't see any retractions, thus wondering if my theory is wrong.

just curious? how did you actually view the results of the query?

if you mean by how do we see the retractions, usually by pushing the results into a kafka stream.