Question

Design a webhook delivery network that reliably delivers event notifications to millions of customer-owned HTTP endpoints. Peak is 500k events/sec, customer endpoints are flaky (timeouts, 5xx, slow, sometimes down for hours), and you must deliver at-least-once with exponential-backoff retries for up to 24h, preserve per-resource ordering where the customer requests it, sign payloads, and never let one slow customer's backlog starve everyone else. Walk through the delivery pipeline, the retry/ordering model, and the central trade-off.