Docker Hub outages and rate limits are a hidden single point of failure in CI pipelines. When Docker Hub goes down (as it did for ~12 hours in October 2025) or returns 429 Too Many Requests errors, builds fail regardless of your paid plan tier. The recommended fix from Docker, AWS, GitLab, and Red Hat is a read-through cache in front of the registry. A comparison of CI platform approaches shows that GitHub Actions and CircleCI only offer rate-limit exemptions (not true caches), while Blacksmith, Depot, and Nx Cloud provide actual caches that survive outages. Nx Cloud's Docker read-through cache is presented as an add-on for teams already using Nx Agents on dedicated compute clusters, requiring no changes to Dockerfiles or image references.

6m read timeFrom nx.dev
Post cover image
Table of contents
Your build stands on a single point of failureEither way, the vendors agree: cache itWhat about the CI platforms?The Docker read-through cache in Nx CloudStop pulling straight from Docker Hub
419 Impressions