That’s because you stop receiving queries. Your conntrack table (or equivalent) is empty and thus following NTP queries from the monitors work.
When it’s in the pool, clients connect to it, thus saturating the conntrack table, thus dropping the extra queries (including the monitors).
Update: I solved the problem by decreasing net speed
connection tracking limits seems to be quite a common problem. the simplest solution is to reduce Net Speed to whatever your router will handle.
it’s bizarre there’s not a straightforward router configuration change to allow routers to stop connection tracking:
It depends on your router - for plain Linux, or for something like a mikrotik box, this is very easy to achieve (ed: provided you don’t need to also NAT the untracked packets).
For a typical home router though, there’s really no point. Normal home users don’t put anything like the kind of state table load on their device to warrant implementing that feature. Running an NTP pool server is very much an outlier here; there is very little else that somebody might want to run on a home connection that has such a diverse UDP client base.