Envoy's 15 s route timeout vs a 30 s Sume sync call

Why does a gateway cut a sync call at 15 seconds?

Envoy's route configuration says RouteAction.timeout defaults to 15 seconds, 0 disables it, and the value includes retries. idle_timeout is a separate setting. Sume's sync mode waits at most 30 seconds before it returns a job envelope, so an Envoy route with defaults can give the client a 504 first.

What should I change?

The cleanest fix is not to wait. Leave mode unset so Sume answers with a 202 envelope in well under a second, then follow status_url from the client or take a webhook. If you must wait synchronously, set the route timeout above 30 seconds and account for retries: a retry inside the route window can replay your POST, so keep the Idempotency-Key stable.

Setting the timeout to 0 removes the safety net for every request on that route, so prefer a specific longer value.

routes:
- match: { prefix: "/v1/avatar-1.0/" }
  route:
    cluster: sume_api
    timeout: 45s

What does the client do on a timeout?

Do not send the paid POST again. Poll the job you already created; the jobs docs say not to resubmit the original paid request only because a local process timed out. The Heroku variant of the same problem is in the H12 post.