When you use: - Postgres source with
xmin incremental replication (not CDC), and - Destination sync mode
Incremental | Append + Deduped to Snowflake, **deletions in Postgres are
not propagated to Snowflake.** Reason: - With xmin, Airbyte only re-reads rows that appear in the source query (i.e., rows that still exist and have changed); deleted rows no longer appear, so the connector has no signal that they were deleted. The Postgres docs explicitly call out that the xmin method “cannot support row deletions.” [[Postgres troubleshooting](
https://docs.airbyte.com/integrations/sources/postgres/postgres-troubleshooting#connector-limitations)] - For standard incremental syncs (non‑CDC), Airbyte’s own tutorial states that
delete operations are not propagated; the destination tables (raw and normalized) remain unchanged when a row is deleted in the source. [[Incremental tutorial](
https://airbyte.com/tutorials/incremental-data-synchronization#step-7)] So in your setup: - The row is deleted in Postgres. - It simply
remains present in Snowflake in both the raw and the deduped (“final”) table; no soft‑delete flag is added and no hard delete occurs. If you need deletions reflected in Snowflake, the documented options are: - Use
CDC (log-based replication) instead of xmin; CDC can emit delete records, which destinations can use to remove/mark rows. [[Postgres troubleshooting](
https://docs.airbyte.com/integrations/sources/postgres/postgres-troubleshooting#connector-limitations)] - Or use
Full Refresh | Overwrite to fully mirror the source, at the cost of higher load. [[Forum: deleted records](
https://discuss.airbyte.io/t/372)]