Millions of Facets

["Hicks, Matt" <ma...@outr.com>]

I'm writing a messaging platform and indexing the messages in Lucene. In
order to determine what messages a user is "subscribed to", I add their id
as a "subscribed" facet.  This is problematic as we scale up since it means
that each message could potentially have millions of subscribers for the
public streams.  Is there a better way to handle this that will scale
better?