Connectivity & power window — 03:15–03:35 PM (EEST)

Reading

This 20-minute stretch is noisier than the device's baseline: 11 brief offline→online flaps occur, each a few seconds to ~38 seconds long. In 10 of the 11, current resumes right where it left off (1300–1700 mA) — those are ordinary MQTT/Wi-Fi reconnects, not evidence the load changed, and none came close to the 5-minute streakBreakGapMs or 30-minute confirmMs the freeze detector uses.

One gap is different. At 12:27:24 UTC (03:27:24 PM), after a 37 s gap, current reads 0.24 mA then 0.49 mA while voltage jumps to 5.15 V (vs. the nominal ~5.09–5.10 V) — the load actually disappeared, not just the telemetry. It's followed by a ragged restart ramp (456.8 → 862.5 → 1346.7 mA across ~5 s), a brief undershoot around 12:27:44–53 (1032–1187 mA), and then a settle into a visibly higher steady regime (~1650–1850 mA) than the ~1350–1650 mA cycling seen beforehand — shaded in the lighter red band on the chart.

This is the same brownout/reboot signature seen at the onset of the earlier, much longer freeze event (idle period 3), just far shorter and occurring mid-activity rather than after a long idle. No freeze_suspected event was logged for this window, correctly — the detector is scoped to sustained freezes, not brief reboot blips like this one.