Timebound + reward based and slashing ( punish bad behaviour and incentivices good behaviour)
Are key ingredient to have a powerful spam avoidance mechanism.
Cycle cost might be a good reference value, but I wouldn’t bound it to usd. Important is, that the cost to spam should always be last attractive.
Another possible is to allow the stakers to decide whether the initial cost for a proposal shall be adjusted or not. We should not bound it to usd. Ethereum doesnt do it either regarding the required number of eth to run a node.