Agreed, but we have to consider the pools as attackers first sadly.
Any staking pool operator controlling more than 20% of global stake puts the network at risk.
Mitigation ideas already in the paper:
1.To make nodes as accessible (low cost) and simple to use for individuals. (one click nodes) (At launch). Mobile OS run node/wallets could be a pretty effective mitigation method.
2.Finding the lowest staking cap (with a reasonable max carbon footprint) that allows maximum individual participation vs pool (At launch).
3.Punishing bad behavior disconnecting offline Algos from staking and making them pay to re-enter.
For your consideration, 4.
Maybe we could limit the activation staking rewards to the range of a healthy Algorand Network:
A block reward multiplier derived from the node successful hit rate of the block round (n) to be applied and paid in block (n+1) where:
80% successful node hit rate, sets the multiplier to 0 and therefore, no fees, no party, bad pools.
100% successful node response, sets the multiplier to 1 and full rewards are distributed to everyone. (plus you can throw some $Oranges in everyone’s bags for kicks and giggles)
In this case is in the best economic incentive of the Pools is not only to concentrate as much control of the stake to increase the size of the rewards received but to always keep the online stake rate at its highest and healthiest to fully materialize those rewards.
All these strategies would never fundamentally defend the network from a non-economical (pure-evil) Attacker, because at large global infestation numbers of cancerous Algos the VRF becomes comfortable and very predictable for the attacker. Maybe introducing oscillation through a feedback loop in the VRF could help but that’s for some other day.
Much 
to all, Algorand!