sometimes an active agent might not auto-register as expected. this is hard to debug if there's no access to the agent and the server handles a lot of agents - increasing log level for all trappers is not feasible.
currently, the best bet seems to be tcpdump by ip, which is not that convenient.
it would be great if debugging could be tuned by the peer ip address - for example, "act like debuglevel 4 if the ip is nnn".
this is an even more fine grained control than sought in ZBXNEXT-4075.