Looking at the server-status on lighttpd, running behind a AWS load balancer, I'm seeing some long-lived (up to around 60 seconds) requests, in a "read" state with a blank host and blank URI. What does this mean, and is this anything to be concerned about?
It seems that ELBs keep TCP connections open to the back-end instances and pass requests through that TCP connection. This lowers latency for requests because they don't have to wait for a new connection to be opened up. The number of open connections scales up and down with the number of requests the ELBs are passing through to the back-end instances.
It can potentially cause problems if you have multiple ELBs pointing to the same instance and you have a limited number of allowed connections as one ELB can be holding a connection open while another is not able to make a new connection. This situation is, however, a bit of an edge case and is probably an indication that you need more capacity.