Hello, @DZack23! Thank you for the reply, very clear answer 
I understand! Mind if I ask further questions? I’m still a bit lost on a few details haha.
To simplify my comment, I’ll describe as if there were a single honest validator (the “defender”, who’ll post the single honest top-level edge in each challenge) and a single dishonest validator (the “attacker”, who’ll try to mess things up). Take that to generally mean “the set of honest validators” and “a set of dishonest validators working in tandem” respectively. Alternatively, take that as pessimistic scenario where there’s only a single honest validator defending the rollup against the world.
If I understood it correctly, in a multi-level setup the defender can be forced to “match” the number of stakes submitted by the attacker on the previous challenge level. The mitigating factor is that nested challenges have a smaller stake.
I’m curious about this mini-stake. Can you give more details, or point somewhere I could read more about it? I’m wondering how small it is. Do they get exponentially smaller as nestedness increases? I guess not, otherwise either the confiscated stake wouldn’t be able to reimburse the costs, or the first stake would have to be very high.
This is because invalid sub-challenges don’t buy any additional delay, they only slightly grief the honest team.
I don’t get this 
What buys additional delay, and why sub-challenges don’t? My understanding is that the height of the history can be as large as the parent one, which would grief just as much.
You can see that I’m a bit confused haha, some detail probably went right over my head. Thank you again for the reply.