What are best (advanced) practices for scaling Facebook AEO campaigns?

Eric has covered most aspects of this. What I'd add to this is to experiment with different audiences as well, which IMO can be crucial in helping you reach users that you may not have reached before.

We've found success in experimenting heavily with lookalikes of users who've completed different upfunnel events(completed tutorial, completed level X - in addition to lookalikes of purchasers). Ditto with experimenting extensively with broad & interest based targeting.

AEO (and VO) bid types are funny, because Facebook rolled them out without much guidance on how to make them work properly. And what I've seen is that results differ dramatically for different strategies across different app types. I'm not sure why Facebook hasn't published more material on AEO and VO best practices given how important those mechanics have become to most app advertisers' acquisition efforts.

In any case, it seems like you are asking for very specific instructions for improving the performance of your AEO campaigns, which unfortunately I don't think anyone can give. In my experience, the key to making AEO work is rapid experimentation: with audiences, with creative concepts, and with in-app events. You need to constantly be spending budget against combinations of these things to find the right configuration that 1) reaches a large monetizable audience at low cost and 2) sends Facebook signal fast enough to allow it to optimize its targeting within that audience properly.

So some tips:

1) Experiment with up-funnel (early stage) events. Try to see how early in the user lifecycle you can go to still produce valid signal; can you find something in the first session that is an indicator of monetization? Keep exploring different events until you find something;

2) Make sure you are actually experimenting with different creative concepts and not just variants of the same concept. More on that here: What's the best way to test creatives on Facebook?

3) Avoid Facebook's A/B testing tool. A/B testing is slow, expensive, and it doesn't fit the AEO paradigm, which is more of a Bayesian optimization mechanic. There's no point in knowing which creatives work best for a static audience at static spend when Facebook will be dynamically optimizing an ad set at scale.

4) Try to avoid audience overlap. This is difficult and may require you to set up new ad accounts or run tests in a sequence as opposed to in parallel. 

I hope that helps: I realize that these aren't very specific instructions, but you won't get valid marching orders that apply to your app from anyone.