n8n Wait node under 65 seconds: how to poll a Sume job

Poll status_url in a loop, wait next_poll_after_seconds between checks, and know which side of 65 seconds the wait lands on. The n8n page says waits under 65 seconds do not offload execution data to the database, and longer waits do.

Sume behavior is from the Jobs and results docs; the n8n text was read 2026-09-30.

What changes at 65 seconds?

n8n states: “For wait times less than 65 seconds, the workflow doesn't offload execution data to the database.” Longer waits persist it. Its resume modes include After Time Interval and On Webhook Call.

Where does the wait length come from?

The async 202 and status responses can carry next_poll_after_seconds. Feed it to the Wait node instead of a fixed number, and fall back to your own interval when the field is absent. Stop the loop on a terminal status: completed, failed or canceled.

curl https://api.sume.com/v1/jobs/job_123/status \
  -H "Authorization: Bearer $SUME_API_KEY"

When is a webhook resume the better fit?

For jobs that run long, the On Webhook Call mode means no polling at all: submit with a public HTTPS webhook_url and let the signed job.completed, job.failed or job.canceled event resume the run. Verify the signature before you trust the body.

Does a failed poll mean the job failed?

No. A timeout or error on the polling request says nothing about the job, which keeps running. Poll again and never resubmit the paid request.