I have another viewpoint, and that might be that perhaps the millions of facts don't originate on this machine, and hence, the places where the facts originate

I have another viewpoint, and that might be that perhaps the millions of facts don't originate on this machine, and hence, the places where the facts originate could have a SPARQL query node... any other query node could then aggregate the results, and you would then in effect have a distributed cluster.