Record identity: lab result IRIs change from every import source;
conditions, allergies, immunizations and patient profiles change from
FHIR sources; every record type read from a C-CDA document changes.
Records the source never identified now get a content-derived IRI
instead of a random one, so re-imports reconcile.

Data recovery: a repeatable C-CDA element had two shapes and four
sections read the wrong one, so labs, vitals, family history and
implanted devices imported as zero records from a vendor-recognized
document. Allergies in the standard nesting produced nothing. Procedure
records were written with no name, date or code, and a resolved problem
imported as active. An import now reports entries read against records
imported per section, and names any section that yields nothing.