Some more on TIME_WAIT continuing from http://www.port80software.com/200ok/archive/2004/12/13/212.aspx...
For some additional examples of TIME_WAIT and port exhaustion woes in an HTTP benchmarking context, check these out:
http://mail-archives.apache.org/mod_mbox/httpd-dev/200009.mbox/%3C39CA6854.1904EFBC@raleigh.ibm.com%3E
http://www.developerweb.net/forum/archive/index.php/t-3597.html
http://codeguru.earthweb.net/forum/printthread.php?t=325935
Ironically, because of the peculiarities of a stress testing environment, the problem is often manifested on the client side, but this can be overcome a combination of good test tool design and an adequate number of physical clients.