[...] On the other hand, if you see that there have been two or three of those spikes reaching 7-10k vb/s you know that at least for a few hours there is little chance of another set of waves hitting the mempool.
Correct me if I'm wrong, but in the case of a so-called burst of transactions, they are not divided evenly between blocks? I don't remember who exactly published it, but a couple of months ago I saw a test here on the forum that simulated 300 thousand transactions, which were ultimately evenly distributed between three or four blocks.