How to Track Connection Success and Drop Rates
Connection success answers, “Can you use the service after tapping connect?” Drop rate answers, “Once connected, does the session drop unexpectedly?” Before testing, choose a target app, a connection timeout, and a session observation window. A successful connection means more than seeing an icon light up in the client: confirm that the target app can complete an agreed task over the route being tested, such as loading an uncached page or sending a real request.
| Metric | How to measure it | Use case |
|---|---|---|
| Connection success rate | Attempts that complete the target task within the timeout ÷ total connection attempts. | Quick research, or reconnecting after switching networks. |
| Session drop rate | Sessions with at least one unexpected drop during the observation window ÷ established sessions that completed the window. | Meetings, remote desktop, long transfers. |
| Recovery time | Time from confirming a connection interruption to the target app completing the agreed task again. | Checking whether work can resume after reconnecting. |
If a session drops and recovers several times, count it only once in the drop-rate numerator, but record each interruption separately. Keep user-initiated exits or sessions ended early for external reasons as a separate group; don't mix them into completed observation windows. If a route never connects, its drop rate is not calculable—not zero. Keep both numerator and denominator when comparing results, and don't treat a small sample with no drops as a reliability guarantee.
Direct Connections, Relays, and IEPL Lines
A direct connection usually means the client connects straight to the destination node, without a separate relay entry point arranged by the provider. It does not mean packets avoid routers on the public internet. There are fewer steps, but performance depends on the public route from your local carrier to the node. A relay connects to an entry point first, then the provider forwards traffic to the exit. The path between entry and exit may suit your carrier better, or traffic may queue at a shared entry point, so the “relay” label alone says little.
IEPL is an international Ethernet private line. If a node is marked IEPL, check which part of the route the dedicated line covers. Even when one segment uses a private line, the access network between your device and the entry point, or the route from the exit to the destination site, may still be a bottleneck. To check route regions and types, start with the server routes page, then test on your actual network rather than treating a topology label as a result.
Protocols also affect performance. Hysteria2, for example, is based on QUIC and uses UDP; if your access network restricts UDP, problems can arise from the start of the connection. VLESS can use different transport methods, so its name alone does not tell you what the underlying route uses. When comparing protocols, keep the region, node, and destination fixed. If changing protocols in the client also changes the entry point, treat the result as a comparison of complete route configurations, not of the protocol alone.
Peak-Hour Stability Test Steps
- Keep conditions consistent. Use the same device, access network, and client version. Record the routing mode, route name, and protocol. Turn off options that automatically switch nodes during the test, so backup routes don't affect the results. Keep the device awake when testing meetings; run a separate set of tests for background performance with the screen locked.
- Alternate connection tests. Rotate through candidate routes during the same time period. End the previous connection before each test, then verify the new connection with the same task. Record handshake failures, name-resolution failures, and app request failures separately. Automatic client retries count within the current timeout window; don't count each retry as a separate attempt.
- Observe sustained use. Run the same-length task on every route. For meetings, watch for audio or video interruptions; for remote desktop, note long periods without a response; for file transfers, check whether they resume after an interruption. Record the time of each drop, whether the client reconnects, and when the app recovers. The latter two may differ.
- Retest at typical usage times. Record off-peak and peak-hour results separately, and repeat tests on different days where possible. Don't directly rank one route tested during the day against another tested at night. If you use both home broadband and a mobile network, draw separate conclusions instead of combining the two access conditions into one score.
Distinguishing Drops, Congestion, and App Timeouts
When performance stalls, first check whether the operating system switched access networks, then compare the client's connection status with the target app's logs. If local access is also unstable, start by checking Wi-Fi or broadband. If the client still shows a connection but the target site won't load, try a comparison destination confirmed to use the same route. If only the original destination fails, check that site's status, DNS results, or access policies. If several destinations fail at once, investigate the entry point, exit, and protocol. These symptoms help narrow the search, but they don't identify the cause on their own.
If requests repeatedly slow down, latency fluctuates, or performance stutters during peak hours while off-peak use is normal, congestion is worth investigating first. Also check whether there is heavy local uploading. A lit connection icon doesn't mean app traffic has not been interrupted; conversely, one app timeout doesn't necessarily mean the tunnel dropped. Track “connection interrupted” and “app unavailable” separately, then summarize using consistent criteria. When sending logs to support, include the time of the issue, route name, and error type, and remove subscription links and credentials.
Choosing a Primary and Backup Route
- ✅ For brief browsing, compare connection success and time to first usable response, then check responsiveness during use.
- ✅ For long sessions, focus on drop rate and recovery time. If reconnecting still requires rejoining a meeting, count it as a real interruption.
- ✅ Once you choose a primary route, test a backup with a different entry point or path where possible to reduce the chance that both routes are affected by the same failure.
- ❌ Don't combine results from different carriers or observation windows, or mistake a one-off speed result for long-term stability.
Choosing a route: Test with the same real-world task on your usual network during your busiest hours. For short tasks, favor the route that connects more consistently; for longer tasks, favor the one with fewer interruptions and faster recovery. If results are close, keep observing. There's no need to replace a reliable primary route just to get a higher peak speed-test result.