I’ve started recording the contents of inbound and outbound getblocktxn messages a week ago. This should allow for some insights into “are peers often missing the same transactions?” and “can we pre-fill the transactions we had to request our self?”. I haven’t taken a closer look at the data yet.
Also, I’ve changed one of my nodes to run with blockreconstructionextratxn=10000 and updated two nodes to a master that includes p2p: track and use all potential peers for orphan resolution #31397. Probably need to wait until the mempool fills up again to see the effects of this.