Cloudflare experienced another significant outage on December 5, just two weeks after a previous one that affected a large portion of the internet. This recent incident lasted 25 minutes and impacted about 28% of its HTTP traffic. The root cause was a global configuration change made during the rollout of a security fix for a React vulnerability. The fix triggered an error in internal testing tools, leading Cloudflare to disable those tools via a global killswitch. Unfortunately, this action inadvertently caused HTTP 500 errors across its network. The recurrence of such outages raises concerns about Cloudflare's reliability. After the November outage, the company committed to hardening its configuration processes, but the necessary changes have not been implemented yet. Customers may find it unacceptable to face similar issues so soon. Cloudflare's reputation hinges on its reliability, and repeated failures can drive customers to consider backup CDNs. In its postmortem, Cloudflare’s CTO, Dane Knecht, emphasized the need for safer configuration rollout strategies. These include enhanced rollouts and versioning, streamlined emergency capabilities, and improved error handling to prevent services from dropping requests during failures. The article also highlights a broader issue in the tech industry: global configuration errors have historically led to major outages. Examples include Meta's 2021 DNS-related outage and Datadog's 2023 incident caused by simultaneous OS updates. Implementing safer rollout practices may be essential for large systems, but the trade-offs in development speed can be a concern for smaller companies.