kane is the tor/i2p only node. kane’s plot is interesting - all nodes except kane have fresher new table (more recent timestamps) and staler/older tried table. so maybe kane hasn’t been receiving fresh ADDR gossip but is still actively connecting to peers which keeps tried table fresh?
also compared alice’s current addrman with a few older snapshots I had (last week, nov 2025, aug and sep 2024). it’s interesting that the older snapshots had lots more 2 week - 1 month old addresses in the new table compared to the recent snapshots.
I did cross-check claude’s script that it was computing the buckets (<1h, 1-6h, 6-24h, .. etc..) correctly from each snapshot’s own reference time - basically the newest entry in the addrman is chosen as the reference.
sorry, this was incorrect. it’s expected behaviour in tor/i2p-only node’s addrman that tor addresses can at max occupy 1/4th new table and i2p addresses another 1/4th. so that’s why we’re seeing only 1/2 the new table being fresh.
if we do the netgroup calculation - tor/i2p addresses have only 16 possible netgroups and only 256 buckets possible in new table (256/1024 = 1/4 new table occupied). see this writeup in case anyone’s interested.
