So I had 5 clones, and I tried to promote each one to a different class. In the first few upgrades they are so similar I didn't notice what was supposed to be different about them from watching them move and interact with the enemies. But nothing a bit of rebalancing wouldn't fix.
You can make the first few upgrades have a "great" effect, then have them get exponentially more expensive, and diminish their returns (i.e. you pay 10x the price to get 1/10th of the upgrade). That way, in the beginning, you really feel the upgrades stacking, but as you get more advanced, you really need to accumulate points to make them go from doing 1000 damage to 1100 damage. See this article.
But yeah. I was in a bit of a rush so I didn't hover over anything, I was clicking. A LOT. lol Maybe something in the UI to visualize what the upgrade (i.e. 100dmg >> 110dmg) would help communicate it.
IDK about locking the promotion. I think clearly communicating the effect on the clone should be enough. In my opinion you don't want to restrict too much player expression. If I want to promote right away, I should be able to do it and be aware of the tradeoffs.
I like the powerup idea. You could even have a special clone that just collects powerups!
As for the anticipating enemy direction thing, I made a `TargetManager` class for Recolonizer, and it's definitely overkill for your project, but who knows, you might find something in here to improve the targeting of the clones. It's where I wrote the predictive targeting logic (see GetCorrectedTargetPosition). It has quite a few dependencies for some other functionality (such as using physics to shoot a rigidbody projectile perfectly on target), so if you want me to share more just ask.
Good luck on your project!