Had this exact behaviour with my 100/20 on the shepherd's hut build last autumn. Turns out it's almost always one of three culprits:
1. Panel voltage dropping below the controller's reconnect threshold
On heavily overcast days, a single 12V nominal panel can sag well below what the MPPT needs to "see" to reconnect. Check your VPanel readings in VictronConnect during the dropout — if it's hovering around your battery voltage plus a few volts, that's your answer.
2. Absorption/float cycling
If your batteries are near full and the sun's patchy, the MPPT will legitimately switch off rather than backfeed. Fogstar Drift LiFePO4 cells are particularly aggressive about rejecting charge near 100% SOC.
3. Wiring resistance
Undersized DC cable between panels and controller causes voltage drop under load, which can trigger the low-panel-voltage cutoff even when irradiance seems adequate.
Worth checking in VictronConnect → History → whether you're seeing actual faults logged or just normal on/off cycling. There's a meaningful difference. A genuine fault will show in the alarm log; normal threshold behaviour won't.
What's your panel configuration and battery chemistry? That'll narrow it down considerably. Also — which specific MPPT model, and have you checked whether the firmware is current? Victron pushed a fix a while back that addressed some spurious dropout behaviour on the 75-series controllers.
Anyone else seen this with partial shading setups? I've wondered whether running two smaller strings into separate MPPTs (rather than one combined string) might reduce this on grey UK days generally.