These are all perfectly reasonable questions, but I think they might reveal a case of Soft Launch Anxiety Syndrome: a condition in which a developer thinks they need to try to control for every single variable in their game's soft launch performance in order to make a final decision on going to global launch or not.
You can't do that -- you'll never truly know. You won't know:
- How acquisition costs will increase as you scale in global launch;
- How cohorts' retention profiles will change after 90 or 120 or 150 days in the game;
- How player spending patterns will evolve over months or years in the game;
You can't know these things, and elaborate spreadsheet models or mental gymnastics around country population ratios and GDP per capita will only just mask that. You can't make good decisions by trying to control market forces that are uncontrollable to pinpoint some specific number (CAC at scale, size of TAM, etc.). The way to approach a soft launch, fundamentally and philisophically, is to try to control risk.
I wrote about how I think about soft launches in this answer: What are some of the best practices for soft launching an app?. To address your specific questions within the context of that framework, I think you should be asking yourself: "Who is the core audience of our game and how can we best reach them in a way that is representative of how we'll do that in global?".
Your first question is really only relevant in a cold start situation, but in that case you can target broadly. Is your game more male or female focused? Start there, and use comparable games for interest targeting. Note that I believe in a soft launch, you should be relying on Facebook for the vast majority (definitely more than half, perhaps more like 75-90%) of your traffic. Why? Because Facebook will be the majority of your spend at global scale, and you have to prove that it works. If you can't scale your game efficiently on Facebook, it won't be successful. So Facebook should be the primary channel you use in Soft Launch.
And since that's the case, you really only need to "guess" who your audience is once (although you should intuitively know this anyway -- who did you build the game for?), which is when you kick off campaigns. After that, your user acquisition should rely on a process that creates / updates custom audiences based on some success metric. Whether those custom audiences are used in event-based (AEO, VO) campaigns or MAI campaigns will depend on performance (for Soft Launch, I don't recommend VO campaigns).
As for budget and CPIs: don't worry about CPIs (you care about scaling user acquisition profitably), and sure, keep the budget constant throughout. But you'll also want to stress test ROAS as you scale spend and broaden your audiences, so it might make sense to meaningfully increase (2x) the budget at some point to see if ROAS holds. You don't need to bend over backwards to maintain CPIs if you're solely focused on your game's ROAS curves.
Regarding iOS vs. Android: this is similar to the channel issue -- which platform will be most important to you? Which platform does the game have to work on in order for your game to be a success? You can run your soft launch across both, of course, but if you're resource constrained, pick the platform that's most important to you.
A note about risk: controlling risk is not achieved by attaining total clarity around the future (impossible). Risk is controlled by limiting the impact of adverse outcomes. You do this with a game launch by scaling systematically: don't worry about CPI, just try to recoup your money as quickly as possible and slowly extend that timeline as you prove out the ROAS curve. Platform featuring isn't meaningfully impactful anymore -- you shouldn't be thinking about a "big" launch for the purposes of getting featured. If it'd be irresponsile for you to put real money at risk for the sake of a big, mass-awareness-generating launch, then just scale slowly and deliberately with short-but-increasing payback windows.