Clash Verge Rev
A node can look selected while your traffic still uses a different route. The usual cause is not a broken node but a different proxy group being referenced by the matching rule. Drawing on common workflows shown in beginner and advanced video tutorials, this guide turns selection and testing into a repeatable diagnostic process. Use only trusted profiles and follow local law and service terms.
Node and proxy group are different layers
A node is a specific outbound endpoint. A proxy group is the decision point used by routing rules. Rules normally target groups such as Node Selection, Auto Select or a provider-defined group. Changing a node in an unrelated group will not affect the current request.
Start from the Rules view: identify the rule matched by the destination, then follow its target to the active group. Rules are evaluated from top to bottom, with an eventual MATCH rule commonly acting as the fallback.
Choose the right strategy
- select: manual choice, useful for a fixed exit or controlled troubleshooting.
- url-test: periodically probes candidates and selects from the measured results.
- fallback: prefers the first healthy candidate and switches when it becomes unavailable.
Display names are chosen by the profile author, so a group may not reveal its underlying type. Inspect the profile structure when labels such as Auto or Failover are ambiguous.
A reliable selection workflow
- Refresh the profile and check for download or parsing errors.
- Use the matched rule to locate the group that actually carries the request.
- Run one health check to remove unreachable or timed-out candidates.
- Test two or three close candidates with the real site, download or application.
- Keep a backup on a different region or route for fast recovery.
Latency is not throughput
A health check mainly measures reachability and round-trip time to its test URL. It does not directly measure bandwidth, peak-hour congestion, packet loss or application compatibility. Treat the lowest number as a shortlist signal, then verify with real traffic. A slightly slower but stable route can perform better than the lowest-latency result.
Common mistakes and safe diagnostics
Avoid changing several global settings at once. Record the chain of destination, matched rule, proxy group and selected node, then change one variable at a time. This separates routing errors from profile, network and endpoint failures.
Never publish subscription URLs, node credentials, private controller ports or unredacted logs. A useful support report needs the client version, time, error category and minimal reproduction steps—not private connection data.