Surviving Meta's rate limits: engineering WhatsApp for 500+ messages/second

The WhatsApp Cloud API will happily return HTTP 429 the moment you exceed your tier's throughput. A naive sender loses messages; a resilient one absorbs the burst.

The architecture is a distributed queue (Kafka or Redis) in front of the send path, with a rate-aware consumer that paces output to the account's current tier. Backpressure is a feature, not a failure.

The other half is inbound: webhook ingestion for delivery receipts at hundreds of thousands per minute, with dead-letter queues and exponential backoff so a downstream hiccup never drops a status update.

This is the part standard marketing agencies can't build. It's also exactly the distributed-systems skill set that transfers directly to any high-throughput event pipeline.