Match Batch Results by Identifier, Not Position
A batch job can return results in an order different from your input list. Joining by row position risks assigning one answer to another document.
Give each synthetic input a unique identifier and preserve it through the Batch API workflow. When the results arrive, join by that identifier and check for missing, duplicated or unexpected IDs before using any answer.
Include one deliberately invalid request in the test batch and verify how its failure is recorded. The successful rows should remain identifiable without silently treating the failed row as an empty success. Save the reconciliation report with the output. This is a data-integrity check your code can perform regardless of how persuasive or well-written the model's individual answers appear.
