Redundancy feels like proof. Two sensors, one answer, and the answer agrees, so the answer is right. The step that gets skipped is asking what would make them disagree. Two nodes of the same model, installed the same week, from the same batch, sitting in the same soil chemistry, share almost every mechanism that could drift them. Electrode aging, temperature history, salinity exposure and firmware rounding are common to both. Agreement between them tests only the mechanisms they do not share, which is a short list. This is why a calibration audit uses a different instrument rather than a second copy of the same one, and why a mesh of identical nodes can pass every internal consistency check while being uniformly wrong. Agreement carries information exactly in proportion to how independent the two error sources are.