Before Fivetran, we used a custom Python data pipeline to extract events data from Pendo and made it available to our users for querying. The event counts were fairly consistent between the Pendo UI/API and table data populated by the custom pipeline

We're now asking our users to migrate over to the Pendo tables populated by Fivetran, but our users are seeing inconsistencies compared to the raw data from the legacy data pipeline. 

Fivetran Support (#97971) confirmed that the Pendo connector is utilizing a set of primary key columns for the feature/page/track event tables, and only unique records based on the PK are inserted. We consider this problematic since event data is being filtered out by the connector.  

Our ask is that Fivetran provide an option on the Pendo connector to limit/disable filtering based on current PK fields. We want to give our users access to all the data returned by the Pendo API, and we'll take responsibility of addressing possible duplicates after the raw data has been loaded by Fivetran