Generally this leads to constantly shifting alliances as one realm gets the upper hand, as was seen in original war island.
Then it becomes a game of diplomacy, which can seem equally arbitrary as noble count. It's in the best interest of the two weakest realms to kill the strongest realm, so the strategy becomes about influencing that perception. You want to appear weaker than you actually are, and the two realms that appear the weakest will want to kill the one that appears to be the strongest.
Once that is done, the strongest competitor has been killed off and the remaining two realms are in the same position they'd be in if there were only two realms to begin with. That is, it becomes primarily about noble count.
Noble count isn't just random. I suspect that Ikalak gets far fewer nobles in part because some players recognize its bad topographic position and don't want to be on the handicapped team. Another influence is the atmosphere of a realm, where realms that are active and fun attract and retain more nobles than realms that don't. Atmosphere doesn't just mean roleplays and banter; having a general who is competent helps augment it. In any case, the noble counts of Taselak and Sandalak have been close to even for awhile, although Taselak's seems to be diminishing as the war comes to a close. I think that when two realms occupy roughly even territories on the map, and when they both meet rough standards for atmosphere, as Taselak and Sandalak do now, they will attain roughly equal totals of nobles.
I like the idea of a two-realm war island. I like it because it diminishes the diplomacy component and therefore increases the importance of everything else. A four-way island would also be a good option, as it could turn into an equal 2vs2 war or a true free for all. Three-way wars would logically end in what I described above, the two weaker realms killing the strongest to optimize their own odds, then fighting a delayed 1vs1 - but with a guilty conscience.