Since releasing the first Project Mercury AMI, I've been talking in general terms about the kind of performance you can get out of this Pressflow/Varnish combination. Over the weekend, I was able to do some more concrete testing, and the results are pretty eye-popping.
For my testing, I used standard small EC2 instances and measured cached page throughput as I added each layer in Drupal's caching stack, and compared that to Mercury. The initial results and technical details are posted on the EC2/AWS group on g.d.o, but this morning I put together a graph. They say a picture is worth a thousand words. I have to agree:
As you can see, Mercury delivers an order of magnitude in page throughput, while maintaining top speeds and keeping load low. Simply put, this is what you want if you're dealing with internet-scale traffic (e.g. digg, drudge, perez). For those who are interested, there are more details and some discussion on the ec2 group post.