Thanks for the interesting solution!

Some points:

• If you need global query view over multiple Prometheus instances (including Prometheus HA pairs), then take a look at Promxy. It just queries the configured Prometheus instances in parallel and then merges the results. It doesn’t use any sidecars, so it should be simpler to setup and operate than Thanos Sidecar <-> Thanos Query scheme.
• It would be great creating similar solutions for other long term storage systems for Prometheus such as Cortex, M3DB or VictoriaMetrics.