Interesting. I'll look into this this afternoon. I'm seeing it consistently handle 2500+ req/sec on my setup, but that's with less than 50 concurrent requests (about 10 concurrent requests via EC2 micro instances...easy to change my testing scripts to 200 I think)