You are right that active stakers would get more rewards initially. In some ways it is the mathematical equivalent of submitting an infinite number of governance proposals every day without submitting any spam proposals, at least in the beginning. That’s why it removes the incentives to spam the NNS with governance proposals. They are no longer needed if the spamming is happening for the purpose of higher voting rewards. Those higher rewards for active neurons currently come at the expense of inactive neurons not voting on governance.
To be clear though, the reason this proposal can be an effective tool is that it incentivizes all neuron owners to actively participate in governance. It would be a very strong incentive. Anyone who is staking can intentionally configure their neurons. If they did so, then ideally we would have 95+% participation in governance topics. Anyone who is getting zero voting rewards will eventually ask questions, educate themselves, and select Followees based on intentional decisions instead of default neuron configurations that lead to a completely passive approach and lack of participation in governance.