Most of the work sits upstream of the warehouse, not downstream. Upstream, you introduce a resolution layer between the event collector and the warehouse, configure the identifiers it uses, set merge rules for any unique IDs, and connect any downstream destinations that want the merge events. Downstream, the dbt models and SQL queries that already run against batch-stitched data keep working, because the warehouse output schema stays familiar. The piece that often needs the most attention is your activation stack, which suddenly has access to fresher data and can be redesigned around it.