Bulk generation and CSV

Re-run a bulk job without duplicating everything

Re-running a whole file after a partial failure is how workspaces end up with thousands of duplicates. Re-submit only the failed rows, and use an idempotency key when driving bulk work through the API so a retry cannot double-create.

Reviewed August 20, 2026

Steps

1. Establish which rows actually failed before doing anything else.
2. Build a new file containing only those rows.
3. Through the API, send an Idempotency-Key with the request so a timeout followed by a retry is not treated as two jobs.
4. Keep the job id — it is what support needs to look at the run.

How to check it worked

The workspace code count rises by the number of previously failed rows, and no destination appears twice.

What usually goes wrong

• Resubmitting the entire file after a partial success.
• Retrying on timeout without an idempotency key.
• Discarding the job id, leaving nothing to investigate with.