How Waiting Room makes queueing decisions on Cloudflare's highly distributed network

What this post added

This post details the evolution of Cloudflare Waiting Room's core queuing mechanism. It explains how Waiting Room, built on Workers and Durable Objects, makes dynamic decisions on when to queue legitimate traffic spikes to protect customer sites. The post elaborates on the challenges of distributed state management and introduces the refined algorithm that uses traffic history and configured limits (total active users, new users per minute) to make low-latency queuing decisions at the worker level, while also discussing the trade-offs and potential side effects of this distributed approach.