Design a CI/CD platform like GitHub Actions that automatically triggers workflow execution when code is pushed to repositories, with real-time monitoring and observability features. The system must handle 10 million repositories with an average of 10 pushes per second and burst traffic scenarios. Workflows are defined as YAML files stored in .github/workflows and consist of jobs that run in isolated environments (containers/VMs). Each job contains steps that execute shell commands or reusable actions. The platform must support workflow-level and job-level concurrency limits, matrix builds (running the same job with different parameters), caching of dependencies and Docker layers, artifact passing between jobs, and both GitHub-hosted (ephemeral) and self-hosted runners. The design should include an event-driven architecture that listens to repository events (push, pull_request, schedule, etc.), a scalable runner pool that autoscales based on queue depth, a DAG-based execution engine that respects job dependencies, and an observability layer for real-time logs and metrics. Emphasize high availability, isolation, security, and cost-efficiency at global scale.