I am also having the same issue.
On the product ntppool:
https://www.pool.ntp.org/scores/45.76.111.149
https://www.pool.ntp.org/scores/2401:c080:1000:4175:5400:2ff:fe32:f445
Server is based in Japan.
IPv6 is solid ; IPv4 keeps oscillating
I was almost pulling my hair out - since all my tests from multiple servers and locations (US, IN, Japan) were 100% positive.
Now I loaded these to the beta pool.
https://web.beta.grundclock.com/scores/45.76.111.149
https://web.beta.grundclock.com/scores/2401:c080:1000:4175:5400:2ff:fe32:f445
These were just put today - so they haven’t yet reached the perfect 20. However the trend is very clearly visible - especially for the IPv4 interface. Newark keeps reporting a timeout - not sure if it is behaving any better than the production pool ?
Any ideas how to improve the situation - due to this issue, I guess the pool effective capacity is much lesser than what it should be? If the transport related issues are not easily solvable; is it an option to tweak the score downgrade that the Newark monitor gives - so that it is not downgraded so aggressively? What is also puzzling is why the transport is OK for some polls, and not the others ?