Infrastructure » website

I expect others to greatly enhance this list, but as a starter ...

report on:
the largest clusters of which we are aware
the largest clusters on which we have tested (including DHO)
the sizes of the clusters in our nightly, long-term, and performance testing

discuss:
practical minimum sizes (for replication and availability)
modeled scalability (and why it should be so good)
measured scalability results
observed recovery time after node additions/failures in large clusters
observed statistical deviation in small clusters
any other issues that might affect cluster size