DongU Kang (@nyanrus@sukhi.f3liz.casa) and I (@hongminhee@hollo.social) have drafted a proposal for announcing temporary suspensions of ActivityPub inbox delivery, primarily for scheduled maintenance. Peers can defer queued and new deliveries until the announced end time or an earlier cancellation, avoiding predictable failed requests during the interruption. This complements 503 Service Unavailable and Retry-After rather than replacing existing retry policies.

A DeliverySuspension object can cover one inbox or all inboxes at an HTTPS origin. Mutual links establish which actor can control delivery suspensions for that target. HTTP Link headers let peers discover a notice they missed, and signed objects can be served from static storage while the application is offline. Peers keep their own limits on suspension duration and queue storage, and resume gradually to avoid overwhelming a recovering server.

https://codeberg.org/fediverse/fep/pulls/954

Feedback welcome, especially on controller verification and how this would fit existing delivery queues and retry policies. Thanks!