maybe limit pool to have active connections to 16-20 another pool for reasonably small time period. after that period some part of active connection canceled and reconnect to other pools. some kind of slow/delayed syncro with all pools in reasonably big time period. but one pool does not have too many active connection if overall pool count is big ?
I think that's a good idea. Maybe we should once every few minutes to randomly tear one of the connections giving others the opportunity to connect. It is difficult to imagine how much we can have pools on restart. On the first launch there was only 4 open ip addresses. I recently revised the log file dnet.log. It turned out that my node received more than 1000 connections at the time the network was launched. Maybe it was an attack, but their addresses were different and it seems they using our protocol. Of course, it was impossible to work. I think if we had about 20 pools and about 50 miners in each, it would be the best option to start.
if we have small connection limits. it should be able to detect cpu threads (it was already done) also with network adapters. so if somebody have high-eng rig or server it can be easier to config the load for them. also some options to LAN connection priorities against public network should be added (it has similarity to p2p networks from my point of view)