Why does my sync sometimes take 25 minutes to extract nothing?
Connector: Postgres source connector
This is expected behavior when there are 0 records in the replication slot. It's necessary to have this timeout because some slots are slow to respond if the WAL gets large. From a WAL consumer (Fivetran) side, a hung replication slot and an idle database appears the same.
If a single change comes through, the sync will finish quickly.
Note: If there are never any writes seen in the process and load bars of the connector, the LSN is likely hung. See this article for more information.