October 2026: Replicate straight into Azure Databricks, plus Grubhub and Salesforce updates
October's release cycle carried replication to more clouds and more sources, and gave Salesforce pipelines finer control over what they replicate. Load reliability across the warehouse destinations took the rest of the month.
Replicate straight into Azure Databricks
The Databricks destination arrived in July for workspaces on AWS, so teams running their lakehouse on Azure still had to route CDC output through an intermediate cloud storage first. The destination supports Azure Databricks, and those pipelines replicate straight into Delta tables in the workspace your analysts already query. The detour drops out of the pipeline, and so does the second copy of the data it left behind. Full setup details are in the documentation.
The Grubhub Source Carries Order and Reporting Data to Your Warehouse
Restaurant and food service teams selling through Grubhub have pulled its numbers out by hand, usually as exports that go stale the day they are downloaded. The Grubhub source replicates order and reporting data from Grubhub's Connect and Reporting APIs into your warehouse on every sync. Grubhub revenue sits next to point-of-sale, inventory, and finance data, so channel comparisons and reconciliation run on current figures instead of last week's export.
Salesforce Pipelines Decide How Deletes and Syncs Behave
Salesforce deletions used to reach the destination one fixed way, and teams that needed deleted records gone from their tables had to clean them up downstream. A deleted-record setting gives you the choice: soft keeps the row and marks it deleted for audit and history, while hard removes it so the table matches Salesforce. Pipelines can also switch between Full Sync and Incremental Sync, so you decide where a full reload is worth its cost instead of living with the default. Our team enables the sync mode choice on request, and the documentation has the full picture.
New Features
- Job Load Sync Method for BigQuery: The BigQuery destination can load data through BigQuery load jobs as an alternative to streaming inserts. Large initial syncs and backfills get a load path built for volume instead of relying on the streaming API alone.
- Automatic PostgreSQL Publications: When the connecting user holds the needed rights, PostgreSQL pipelines create and maintain their own publication. New tables join the publication as the pipeline grows instead of waiting on a DBA to update it by hand. (documentation)
Improvements
- BigQuery destinations move rows larger than 10 MB to a load job automatically, so oversized rows arrive instead of hitting the streaming size limit.
- Salesforce History objects load
OldValueandNewValueas JSON, so values of any field type land in one consistent format. - Database sources keep the initial sync running for every other table when one table fails, so a single problem table no longer holds up the rest.
Fixes
- Databricks destinations keep every row through a restart or resync instead of losing data.
- Databricks loads keep moving after a source schema change instead of stalling.
- PostgreSQL parallel initial syncs release their memory when they finish instead of leaking it over time.
- Redshift and Snowflake merges load a row delivered more than once as a single row instead of a duplicate.
- BigQuery destinations load boolean and numeric values into correctly typed columns.
Tell us how Azure Databricks and the new Salesforce settings fit your pipelines, and which delivery platforms you want covered next.