This is the new tightly coupled run on the mini. It's still in progress, but shown here through block 670,000. Note that in the loosely-coupled run just above, divergence started around 450k and the run was stopped prior to 600k confirmations (at almost 16 hours). In the new run we have reached that same point in less than half the time. This is not a result of increased concurrency, this is a result of shifting the load from download to validation and confirmation, so as to keep each aligned, which dramatically reduces the SSD paging requirement. So it looks like this approach is working.
(edited)