Title: Plasma System Monitor shows incorrect network speed (0 / double values, likely sampling issue) Hi, I’m experiencing an issue with the Plasma System Monitor where the reported network speed is clearly incorrect. The values alternate between 0 and roughly double the expected throughput , resulting in a very “digital” graph (no smooth transitions, just spikes). System Monitor Symptoms Network speed frequently shows: 0 B/s followed by a spike that is roughly 2x the actual value The pattern is consistent: 0 → spike → 0 → spike The graph is not continuous, but jumps sharply between values What I verified 1. Raw kernel data is correct Reading directly from /proc/net/dev shows normal, continuous traffic. 2. Manual measurement is correct I wrote a small script that calculates per-second throughput using byte deltas. This produces realistic, continuous values without spikes. 3. ksystemstats is running and accessible plasma-ksystemstats.service is active and stable DBus service org.kde.ksystemstats1 is available Sensor interface responds and exposes methods like sensorData , subscribe , etc. 4. Issue only appears in Plasma System Monitor The incorrect values only appear in: Plasma System Monitor (graph + widgets) Other tools (script, terminal-based monitors) show correct behavior. Interpretation This looks like a sampling / timing issue in the frontend , not a backend problem. My assumption: Sensor data is delivered asynchronously via DBus The System Monitor expects fixed intervals (e.g. 1s) If a sample is skipped or delayed: one update shows 0 next update accumulates multiple intervals → appears as double throughput So effectively: missing sample → 0 next sample → 2x data normalized to 1s Additional context This setup previously worked correctly on the same system Also works fine on other distributions (e.g. openSUSE with Plasma + Wayland) Appeared after updates (likely Plasma / ksystemstats related) Conclusion The backend (kernel + ksystemstats) appears to be working correctly. The issue seems to be in how Plasma System Monitor: subscribes to sensor data or normalizes values over time Possibly related to missed events or incorrect time delta handling. If anyone has seen similar behavior or knows of a related bug/fix, I’d appreciate pointers.