The hashpower a miner produces and pool accumulates is verifiable, thusly, over the x-axis one can calculate their average number of blocks to consistently represent their percentage of network power. It's unlikely to change much inside the 2-week difficulty adjustment window The tx's we're presumably talking about are enormous tx's, and may originate from a sanctioned country, or entity looking for privacy, or an entity looking to lock in an advertised mempool rate months into the future. Otherwise there's no guarantee this tx will be confirmed in any reasonable amount of time even with an RBF 

mallardshead

bitcoin

A [pro] Marathon Slipstream rant:

Not sure if that is possible or if someone would want it. Because I don't think there isn't a bigger guarantee of time window when you're betting in only one miner, and that miner might take some time to find a block. Then the bidding for the next Block in that miner and it's either a first come first serve policy or bidding like the mempool itself wich makes it worse than broadcast to the network. Then if that miner loses hashrate share, you will have to find another one. I think the market will figure it out it's not worth it. 