That window is thicker at the bottom because someone fitted it that way

Run your hand down the edge of a seventeenth-century window and you can feel it: it is thicker at the bottom than at the top. The explanation you get at school is that glass is an extremely slow liquid and has spent four hundred years oozing towards the floor. The real one is that the glazier fitted it that way, because the disc he cut it from already came out of the workshop uneven.

When old glass is measured in bulk, you find panes thicker at the top and panes thicker down one side. Gravity would not have singled out some and not others.

Window glass used to be made by spinning. The process is called crown glass and in Europe it lasted well into the nineteenth century. The glazier blows a bubble of molten glass, opens it at one end, transfers it to a solid iron rod — the punty — and spins it at full speed. Centrifugal force opens the bubble into a flat disc that can reach a metre and a half across. That disc is not of even thickness: it is fatter near the centre, where it was held by the punty, and thins out towards the rim. Several panes were cut from each disc, and every pane came out with one edge thicker than the other. The glazier almost always set it with the thick edge down, because it beds better in the frame and water runs off it. Almost always: sometimes he set it upside down or on its side, depending on how the cut had fallen. Hence the windows that don't fit the theory.

That bullseye you see in the middle of some old panes is the punty scar. It was the part nobody wanted, sold cheapest because you couldn't see through it. Today it is made on purpose and costs more.

Crown glass also leaves faint concentric curves, which are the marks of the spin, and a ripple that makes the view shift as you walk past the window. Modern glass, floated on a bath of molten tin, shifts nothing.

Nor is glass a liquid. A quartz crystal has its silicon and oxygen atoms in a lattice that repeats; glass has the same atoms and the same bonds, but disordered, the lattice frozen half-finished. That deprives it of a crystal's melting point and leaves it instead with a range, the glass transition temperature, which in ordinary window glass — soda, lime and silica — is around 550 degrees. Above that band the atoms rearrange and the mass turns to paste: that is what the blower exploits. Below it, the time an atom needs to change places goes through the roof.

Edgar Zanotto, a materials engineer at the Federal University of São Carlos, put a figure to that absurdity in 1998, in a calculation published in the American Journal of Physics. At room temperature, the time a cathedral window would need to sag appreciably is on the order of a 1 with twenty-three zeros behind it. The universe has been going for some 13.8 billion years, which is a 1 with ten zeros.

What does move in a stained-glass window is everything else. The lead cames that hold each piece give way under their own weight and under the wind, and the panel bows. Glass does fail slowly too, but by breaking, not by flowing: the microcracks on its surface grow where there is damp and stress, a phenomenon with a name of its own, static fatigue, and a hundred-year-old pane can split on its own. At the Sainte-Chapelle in Paris, the thirteenth-century windows were taken down panel by panel in the restoration that ended in 2015, cleaned, releaded and hung again. The lead is from 2015. The glass is still the glass cut around 1245, with the thick edge where its glazier left it.

Museum cases are full of Egyptian and Roman glass vessels, with walls as thin as paper, two and three thousand years old, standing upright. Not one of them has ever needed its base trimmed.