thanos-io/thanos ? reverse-engineered prompt

Build me a Go based monitoring system that can sit on top of existing Prometheus servers and make them feel like one reliable metrics platform.

I want operators to be able to connect multiple Prometheus instances, including duplicate HA pairs, then query them from one place without seeing double data. It should keep recent metrics fast, store older Prometheus data cheaply in object storage like S3 or Google Cloud Storage, and make old queries faster by compacting and downsampling data over time.

Please structure it as small services that work together, such as a query service, a sidecar for Prometheus, a storage gateway, a receiver for remote write data, and background maintenance jobs. It should have clear command line entry points, sensible config files, Docker support, health checks, metrics about itself, and good docs with examples for Kubernetes style deployments.

Look up current Prometheus and object storage docs online if you need to.