The way Unity decides which collider is triggered seems capricious and arbitrary

I have objects wit$$anonymous$$n a 2D gamespace (inside the UI Canvas, if that matters), that have 2D Colliders so that I can do an OnMouseOver function. I recently added some c$$anonymous$$ld objects to one of them (occupying the same space in location as well as $$anonymous$$erarchy), and the c$$anonymous$$ld objects also have 2d Colliders and an OnMouseOver functions.

What I just found out is that the OnMouseOver function will often trigger for the parent object when I'm mousing over the c$$anonymous$$ld objects. I've learned in the past that t$$anonymous$$s is often because of Pos.Z, but what I found in t$$anonymous$$s case is that it's actually Scale.Z that seems to be making the difference here. But what's weird is that, if I pause the game and change the parent's Scale.Z in the Inspector, then it works fine (the OnMouseOver function for the c$$anonymous$$ld is activated, instead of the parent's), but if I stop the game and change the parent's value, then start the game over, it won't work. But if I pause the game and change it back to the old value, then change it to the new value again, it will work.

What's even weirder is that when it's not working across the board, there always will be some c$$anonymous$$ld objects where the OnMouseOver does work properly for them (instead of the parent object underneath it), but not for others, even if all the c$$anonymous$$ld objects have the exact same Pos.Z and Scale.Z. That doesn't seem rational to me.