Since the network is evolving very slowly, I'd write in a post some thoughts.
I noticed a problem with IGS. The mechanism of dissemination of the available gateways should, with a use as low as possible of traffic, make every node aware at any moment of the address of the n closest gateways. What happens instead is that when a node which was a gateway gets removed, this information is often lost. The result is that the nodes that use those gateways will have soon a not reliable link to the Internet.
A workaround could be to set the number of used gateways to 1, and then make sure that the closest gateway is always on.
It is just a ugly workaround. The bug is critical and should be immediately dealt with.
Nevertheless I'm not going to dedicate time to that immediately. At the moment I am very busy with the porting of the code to Vala, and AFAIK nobody (except me) is using the current version.
Another thought. Based on the first tests conducted, it seems that reaching the Internet with different IP addresses is not a problem for most applications. Obviously except that a TCP connection that has been initiated throughout a certain gateway has to be carried on on that gateway. This is guaranteed by netsukuku.
Nevertheless in theory some applications might be bothered that they reach the Internet with an address that changes very often. We cannot do much for them. The nodes that want to use those applications will have to set the number of used gateways to 1, and live with it.