dbt users on Databricks: do your snapshots handle files that arrive out of order?

If you build SCD2 history with dbt snapshot on top of batch files (daily exports, CDC drops, API extracts), there's a failure mode worth knowing about. Snapshots apply data in the order it arrives. So if Tuesday's file lands after Wednesday's: - with strategy ='check', the late (older) value becomes the "current" row - with strategy ='timestamp', the change is skipped and that period just disappears from history Every run stays green and every test passes. Nothing tells you. We found out when someone asked why an employee's promotion showed up as a demotion. We ended up building dbt-iterative-