I investigated that idea of saving out the problematic data, and ultimately decided it was too complex a solution to be worth it. The data set was large and the connections between all the river nodes needed to be saved out, so it was going to be slow and error prone.

But that's OK, we don't really need that anyway. In 0.5, old worlds should still load OK, and you will even be able to host multiplayer from them, it's just that cross platform play won't be an option for worlds created before 0.5. New worlds will be fine for everything, single player will be fine, and multiplayer on old worlds will be fine as long as you're all playing on the same OS.

So after investigating that, I did the work so it will use the new safer code path for new worlds, I did a fair bit of testing to make sure old worlds all look correct on different platforms, and that new worlds are consistent on both Mac and PC. Then today I've set up a new Ubuntu server as a linux build machine, and have compiled and run the headless server on a remote VM, connected to it and had a little playtest. It worked pretty much flawlessly, it just took a little longer to load in the sapiens and tribes and things.

Next step is to have a proper multiplayer play! I have a couple of kids sick at home, eager to give it a try, so we're going to see how far we can get before hitting a game breaking issue for me to fix!