Benjamin Marvin wrote:

I don't believe this is my problem. The debug and packet captures show all of the accounting packets are replied to within the Response_Window and Max_Request_Time frames. (5-10 seconds being at the extreme high end of response times.)

If the responses are all within "response_window", then the zombie period will never get hit. There's only one place in the code which sets "zombie" and logs the message. It only gets run when a response hasn't been received for "response_window". If there is a response... it doesn't get run. If you want to double-check this, go to src/main/event.c, and look for the function no_response_to_proxied_request(). Add this near the top (i.e. before the reference to zombie) if (request->proxy_reply) { wait_a_bit(request); return; } If the zombie messages go away, then something needs fixing. If they stay, then the home server really is *not* responding within "response_window". Alan DeKok.