Handling OIC transformations that need a lot of Fusion data, without a database?

I'm curious how others have approached this:

Common Use Case: A Vendor or Partner File containing thousands of data rows which doesn't have everything needed to fully construct the FBDI or HDL for batch import, so the transformation has to enrich each record with Fusion data. Think department for an assignment, a manager's details, or the payroll period for a date. With Autonomous AI Database, a staging table usually solves this. But many clients don't have a database in their landscape, and adding one isn't always an option.

For those who've handled this without a database:

• Do you call Fusion REST or SOAP services per record or per batch at runtime?