Hi, #snac author here. As @stefano@bsd.cafe days, it's a common practice in ActivityPub implementations to use retry queues.

Specifically, snac respond differently to 404 and 410 http status messages by not retrying, because they somewhat explicitly say that a resource is no longer available; a 403 status may be a transient error related to RSA keys not being up to date, so it's requeued, and all 5xx (server) messages are also always retried, because they are usually temporarily fallen or overwhelmed servers that deserve another oportunity.

CC: @stefano@bsd.cafe