In case anyone is still watching this, I have stats now on the new
system for a day with traffic on the BIG list.


on the OLD system:
Total: 69,384 deliveries
Highest sustained delivery rate for 1 minute:    20.2/sec at 16:13
Highest sustained delivery rate for 5 minutes:   17.6/sec at
17:23-17:27
Highest traffic in a one-hour period:          41,774 deliveries
(16:45-17:44)

on the NEW system:
Total: 68,604 deliveries
Highest sustained delivery rate for 1 minute:    59.1/sec at 09:10
Highest sustained delivery rate for 5 minutes:   48.5/sec at
09:07-09:11
Highest traffic in a one-hour period:          45,314 deliveries
(09:02-10:01)


 59.1 * 3600 = 212,760 potential hourly capacity
 48.5 * 3600 = 174,600
 45314 / 3600 = 12.587 average per second over busiest hour

so I figure the one-hour number is limited by slow-responding or
temporarily unavailable remote sites, or maybe aggregate bandwidth. And
remember this is with LSMTP using only 40 concurrent outbound threads
from inside a VMWare single-cpu guest machine (but the host machine has
hyper-thread enabled).