Just following up. My submission is a bad solution and shouldn't be implemented. This broke the moment we used %pip to install additional libraries.

I sent our wishes to the Databricks reps we work with, but at this time there doesn't seem to be a good way to support XML. In our case, we added a workflow task (scheduled job) to load these XML documents into a delta table, and work the delta tables as one of the sources in our DLT pipeline.