I wasn't really following the old one, and my advice sounds considerably different from what you're getting, but I do have some insight on these subjects. Very long-winded insight.
--Encourage fishing and other historically accurate Polynesian behaviors. Fishing comes to mind first.
--Preserve realism.
I would be very, very wary of these two. "Realistic" and "Historically accurate" tend to be those sorts of terms that are either there for no good reason or being misused in place of more specific goals.
So, in short, why realistic? What's so good about it? What parts of it are you interested in? Saying you want something to be "realistic" really isn't that different from saying you want it to be "like Star Wars." The only difference is that realism fanboys are less likely to realize they've got a problem.
Similar with "historically accurate." This you've got at least a bit more cause for since it's your theme, but you should still clarify just what "Polynesian" means.
--Reduce complexity. #1 goal here.
A common problem.
The best general advice I can give you is to start simple and differentiate along common mechanics, rather than just throwing everything you can think of together. This goes double if "realism" is a primary design goal.
So for instance, rather than determining how eggs vs fruit should work as a food source individually, thinking about how common fruit is and what shape eggs are and such, it's best to treat them as the same underlying mechanic, then add simple, meaningful differences if desired. Maybe fruit is more bountiful but riskier, or eggs are better but require actually setting up chicken farms. The result should be that everything is easier to follow and perform, and you're not tempted to make stupid decisions because they'll affect all food gathering (or crafting, or whatever) rather than just squeezing cactus juice during high tide.
Beyond that, I notice everyone's listing about ten different things each turn. That implies that you're rolling for ten different things per player per turn, but even if that's not the case you're still processing them. That's no good.
Consider rewarding or at least allowing everyone to clump their workers together more. Instead of building huts, farming food, making spears, picking fruit, hunting clams, butchering turtles, and tending to children each and every turn, what if they built huts and farmed food one turn, then built spears and farmed food another, then butchered turtles and tended the children next? Same results, more or less, but less processing each turn.
As for how to encourage that, not sure. You could just limit everyone to a certain number of actions per turn, or eliminate the utility of creating, say, a single adze, or offer bonuses the more tribe members are involved in something at one time. I'm not entirely clear on your system, so this one's a bit harder for me to offer specifics on.
--Fix balance (see below)
I found that all players chose similar choices with their traits. How could I modify them so there's less "good/bad" choices? I want the tribes or clans or whatever to be a lot different from the start.
Looking over the trait list and starting players, it seems like "traits" are more like embark profiles. In this case, there's probably a lot of similarity for two reasons.
One, animals give you something you can't get later, while canoes and crafts take away something you can make later. There's nothing about how one acquires dogs or chickens if you don't have any to begin with, whereas it's easy to imagine building a canoe at some point in the future.
Two, concrete, obvious bonuses or penalties tend to be more significant than vague or numerical ones. Starting with a breeding pair of dogs gives your tribe access to dogs, for instance, while being isolated... doesn't really cost you much, and starting with fewer people just means you'll be slower at the start. Trying to balance "Your tribe has dogs" with "-4 people" is kind of hard to do; most people will go for the dogs.
So if you want tribes more differentiated, you've got at least two options.
The first is to redo traits to something other than starting gear, like persistent bonuses to something. This might be more complex than you're going for, though.
The second is to limit numeric and yes-or-no benefits to being traded with each other. For instance, instead of spending points, let players decide what one or two animals they want a breeding pair of, then decide separately whether they want to start out with extra knives and no canoes or two canoes but no pots or knives or an extra hat but no pants.
Finally, players tend to try to specialize away from each other- so long as there's not a good reason otherwise- if they're interacting with each other in meaningful ways. You're kind of stuck here at present, because you don't want them having contact immediately, but if they knew that someone else was taking a pair of dogs, someone they'd be able to trade with later, they'd be more willing to take a pair of goats at the start instead.