It's always about incentives. Node runners are still free to determine what's get filtered. Miners are free to determine which transaction they mine.
If enough node runners are fed up with a bloated mempool on their node, they should take action and this in turn will incentivize miners to consider the risk of slower block propagation.
Lynn's point does raise a valid about the cost of transactions that need to disincentivize jpeg's over valid tx's.
In times with low tx volumes, jpeg bloat provides extra income stream for miners. While in times of high tx volumes, jpeg bloaters will think twice if the cost of their monkey pics are worth it.