As an administrator, I would like to query metrics that exist in multiple tenants. For example prometheus_remote_storage_shards - I'd like to see what the typical range is, look for outliers.

Previously discussed at #849 but that refers more to metrics generated inside Cortex.

I see two practical considerations: authorisation to access data from other users/instances, and performance.

Other notes from previous conversations:

• the tentant ID could be exposed as a label…
• suspect the locking strategy inside ingesters might need care
• there is no list of users anywhere - you could use all the users with chunks in the ingesters, but that could miss people who haven’t sent anything in ~15mins
  • for many admin purposes, “active users” is fine.

Also related: #577