When you are delivering millions of message events a day, “mostly reliable” is not good enough. Here is how we built a webhook pipeline that does not drop events — even when a customer endpoint goes down.
Durable queues, not fire-and-forget
Every event is persisted before it is delivered, retried with exponential backoff, and only acknowledged once your endpoint confirms receipt. Nothing is lost between us and you.
Backpressure and replay
Slow consumers do not take down fast ones. Each tenant gets its own delivery lane, and a replay API lets you re-request any window of events after an outage.
ML
Marcus Lin
Co-founder & CTO, Wasparks
Marcus Lin writes about messaging strategy, customer communication, and building businesses that put conversations first.