Base on the trace file you gave me. I found 3 attempt to the http://srx.main.ebayrtm.com, the first attempt the fqdn resolved to 119.46.110.225. the second and the third got resolved to 66.135.211.19, and all 3 get response from the server in a very quick manner, which we could not see the time out traffic in the trace file. That is why I last time mentioned it was hard to replicate the issue, and I replied to you with the report and suggestion from our Taas server which you mentioned you already checked the 1,2 and 4 , that would be the potential cause to this performance issue. 3 is the result which lead to the result, if the surgeQ keep increasing which means the there are packets waiting on the incoming queue waiting for the vserver, if the backend server for some reason is busy handling income packet, and the persistence setting on the vserver still lead the traffic to this backend server, will lead to this result.