The algorithm you describe is exactly what I’m trying to do (so that’s good!). But when calculating the feerate diagram for the post-RBF mempool, you omitted the re-linearization of all the old clusters that are affected by evictions – I don’t think we can do that?