I think that AROS is definitely the most likely to stay afloat - MorphOS has already hit very bad times, and the future of OS4 won't be secured until a lot more people can be convinced to buy into it. The nature of an open-source project is to allow it to continue indefinitely.
However (and this might be a controversial view) I cannot see what will happen to AROS once they've reached the ability to perfectly emulate the OS3.1 API. I personally think that open-source projects are great when there's a defined goal to reach (eg POSIX/Linux, Windows/ReactOS, Windows look'n'feel/KDE) but can reach problems when the time comes to add original functionality; with so many developers, its hard to make a clean break and start making completely novel features. Perhaps they'll prove me wrong, but AmigaOS4, while a lot less likely to be kept alive, offers the best chance of a "future Amiga" IMHO.
