One prometheus instance is a SPOF, and might not even be able to monitor all the instances. This task (likely going to be split into subtasks) is to investigate and build out a HA/scaled-up configuration, possibly using some of those tools:
Note that this task is purely about the ability to query data from multiple sources, data storage is still handled on individual Prometheus nodes. While Thanos Store looks promising, it requires an object store (such as swift) and is considered Future Work.