Moving the larger slow queries from a data source pipeline to a stitch function would allow the building of summary tables from large datasets which would in turn enable faster charts and reduced large database queries. Unfortunately large slow queries often require AllowDiskUse = true. Which is not supported in Stitch.