Perhaps, if squid does get confused (two squids on one IP address) you could run the second instance on a fake IP address/alias but I don't know much about that.
And I know all to much about that. IT is done by creating a new loopback ethernet device, and briging it to your real device. to get an ip addess for this device you send out a dhcp broadcast on it, which loops back, crosses the bridge and enters the real network. (unless your computer is the DHCP server). The exact same thing happens with outbound packets. Inbound packets might not need to cross the brige and loop back. I think linux is smart enough to realize that the addressed ip is the same ip as a device to which it is bridged, and tus insert it directly in the virtual device's incomming queue.
A lot of work, but certainly intersting.[Unfortuantely this knowlege is Windows based, and linux may need some coaxing to accept it]