append

bronze:
  tables:
    events:
      source: { system: crm, object: events }
      strategy: append

systems:
  crm:
    type: csv
    path: data/events.csv

Every run adds the rows of the source to the table. Existing rows are never changed or removed.

Topic	Rule
Rows	Kept as they come: no dedupe, no key needed, no NULL-key drop. A repeated batch is appended again.
Audit columns	The same five as replace, so _execution_id says which run added a row.
Retries	A retried run first deletes the rows of its own _execution_id, so a partial write is not duplicated.
--init	Starts the table over with this run’s rows.
Schema	No schema policy is enforced.
merge options	Not used.

Use it for logs and event data where every delivery is new. If a key can arrive twice, use scd1.