Skip to main content

Retries and Delivery Reliability

A delivery that does not return a 2xx response within the response window is retried on a fixed schedule. This page describes that schedule, the two ways an endpoint can end up disabled, and the recovery tools available to you.

note

The exact retry timings and the response window described in these guides may change.

Retry schedule​

A message is attempted immediately, then after 5 seconds, 5 minutes, 30 minutes, 2 hours, 5 hours, 10 hours, and a further 10 hours. Each interval is measured from the failure of the preceding attempt, not from the first attempt.

For example, a message that fails three times before succeeding is delivered roughly 35 minutes and 5 seconds after the first attempt.

Once the schedule is exhausted, the message is marked as failed for that endpoint. It is not attempted again automatically — recover it using the tools below.

When an endpoint is disabled​

There are two separate reasons an endpoint stops receiving deliveries. They are not the same thing and they are resolved differently.

Sustained delivery failure. If an endpoint keeps failing for roughly 5 days, it is disabled. Fix the endpoint, then re-enable it in the webhook settings portal and recover the messages that failed while it was down.

A billing plan that no longer includes webhooks. If your plan is downgraded to one without webhook access, your endpoints are disabled rather than deleted. Their URLs, event subscriptions and signing secrets are preserved, and the change is reversed when you upgrade to a plan that includes webhooks again — you do not have to reconfigure anything.

Recovering from failures​

Open the webhook settings portal — merchants from Settings > Webhooks, marketplace affiliates from "Your Name" > Edit Your Profile — where you can:

  • Inspect the delivery attempts for a message, including the response your endpoint returned.
  • Retry a single message against an endpoint.
  • Recover all failed messages for an endpoint from a date you choose.
  • Replay messages that were never attempted against an endpoint, which is what you want after adding an endpoint or subscribing it to an event type you had not selected before.

These are portal actions. They are not exposed as Refersion API operations.