Infrastructure as a single point of failure: Cloudflare edition
Two separate Cloudflare threads surfaced today. The Quick Tunnels thread flagged high latency variance (30-50ms to EC2 ballooning to 115-750ms) and uncomfortable terms of use. The Spain Docker pull failure thread revealed that LaLiga's football streaming blocks are hitting Cloudflare R2 wholesale, taking Docker Hub down as collateral damage for developers in Spain during match times.
This is a recurring theme on HN with real teeth. When a significant fraction of internet traffic routes through Cloudflare, a single Spanish court order about football streaming breaks Docker CI pipelines for an entire country. The comment 'Cloudflare is cancer. And the tumor is now too big' is hyperbolic but points at a real architectural risk: centralization creates blast radius.
The Cloudflare Quick Tunnels latency numbers are a separate concern for builders using tunnels as a cheap alternative to proper networking: the variance is high enough to break real-time applications and give users inconsistent experiences.
So what?
If your CI/CD pipeline or any critical development workflow goes through Cloudflare-hosted services, you have a dependency on geopolitical and legal decisions made in countries where you don't operate. Diversify your tooling across providers, and test your fallback paths before you need them during a production incident.