If every step we climb above 1sat/b means everybody has to store a backlog of 120MB of transactions in their local mempool will that scale? Maybe having alternative ways of spreading transactions that one wants to see included in a block will be beneficial.
Discussion
Doesn't that imply that you try to have *all* the tx in your mempool and not the ones with current market fee rate?
It scales just fine because transactions that aren't going to be mined in a reasonable amount of time don't need to get into the mempool at all. This is precisely why there is a mempool size limit, along with code that automatically adjusts the minimum relay fee based on what transactions got ejected due to that size limit.