Runway INTERNAL or null failureCode: how long to delay a retry

Runway's page says a task that fails with INTERNAL, a null or missing failure code, or INPUT_PREPROCESSING.INTERNAL may be retried "but you should add a delay", and it gives no number. Sume puts the number in the error: retry_after_seconds says how long to wait, and retryable says whether resending the same request can succeed at all.

Runway text from its Dev docs; Sume text from Errors and spend and Errors and rate limits, all read 2026-10-01.

What does Runway leave to you?

The delay length, the number of attempts, and whether a null code deserves a retry at all. So a client needs its own backoff policy; a fixed guess may be too short in an outage and wasteful in a one-off glitch.

What does a Sume error give you?

The error envelope has a stable code, and the fields below. Log message, but never match on it.

Which Sume failures say not to retry?

For Format runs, a 4xx at create means nothing ran and nothing was charged, so fix the call. Among job categories, validation, auth and quota call for a fix, not a retry; internal says to inspect events and contact support with the request or job id. runtime_unavailable says retry later without aggressive retries.

What is a safe retry loop?

Back off on 429, wait retry_after_seconds when it is set, and send the same Idempotency-Key when resending a submit, since the docs warn against retrying unsafe submits without one. Stop after a small fixed number of attempts and surface the request_id. Library recipes: BullMQ retries and axios-retry for POST.