Latency tests generally reach only one probe address. Real access also depends on the destination, DNS, rule matching, congestion, bandwidth, and protocol transport. Switch between two nodes in the same policy group and retest a download or playback. If only one domain is slow, adjust that domain's rule or DNS instead of changing the global configuration.
The probe is fast; where is the real destination slowing down?
Send a request from the actual slow website or app and inspect its policy, node, protocol, and DNS result. Do not use the latency probe as a substitute for the real destination.
- Is the latency-test URL similar to the actual destination?
- Are all websites slow, or only one website or app?
- Does the destination select the expected policy group in the connection list?
- Were DNS, IPv6, QUIC, or UDP changed recently?(See DNS FAQ)
Compare rules and nodes with the real website or app
- Use the actual domain to inspect connection-list entries and rule matches(See Rule-based Routing FAQ)
- Temporarily compare different nodes in the same policy group
- Disable the browser's independent DoH or proxy extension and test again
- For streaming, downloads, or conferencing apps, inspect UDP, QUIC, and rules separately
- Record the time period, destination domain, node name, and redacted logs before reporting the issue
Low latency does not mean high bandwidth
Probe results are easily affected by the test URL and cache
Do not judge every node from one speed test