Should we change how CSV headers are handled? See discussion in past PR: https://github.com/CDRH/datura/pull/244#pullrequestreview-3073073002