Long running tests of SFSB replication under heavy load.

Should include:

1) Letting a significant percentage of sessions passivate and later activate.
2) Letting a significant percentage of sessions expire due to time out.
3) Invalidating remaining sessions after a while (i.e. don't use any sessions for the life of test; all get used and replaced after a while).
4) Perhaps, a low (say 1/1000) probability of failover.