This check is quite futile and only causes issues with NATs. proxies and other such setups where addresses are expected to change. You should disable that check. A better way to protect your cookies from being stolen is to use SSL/https. The problem is that with the advent of Happy Eyeballs it is really hard to determine what part of the full stack (Kernel, libc, system-resolver, application) makes the decision on which address will be used. Yes, indeed, the application can make their own decision too, which adds to the fun. (OSX is the best in that area as they use latency to determine which address gets contacted, thus then it really becomes random what gets used). AICCU does not make decisions on what addresses gets used, it just sends packets. Now of course, those tunneled packets might be slower than native IPv4 packets due to the tunnel overhead and thus in that case IPv4 might get preferred over IPv6 if something in the stack decides based on that merit. tcpdump could help there, but as you are on Linux just do a 'ip monitor" to see netlink messages that might affect any routing changes. Because of the stack that can decide which remote address is chosen you will have a hard time.... Chromium btw exposes a little bit of it's decision making process, check chrome://dns for this.