Elke had replaced the glass in the east transept three times.
Not because it broke, though it had, twice. The first replacement was storm damage — a branch through the lancet window that took out four panels. Standard restoration: match the original glass, reset the leading, fill the gaps. The second was thermal stress — a cold snap found the flaw line in a panel she’d set herself. That one stung. The crack followed the pattern she should have anticipated.
The third replacement was harder to justify.
The panels were intact. The leading was sound. The glass she’d set after the thermal failure was good work — matched the original’s amber tint, held the same wobble in the surface that gave medieval glass its warmth. Kessler, the preservation officer, had signed off. Technically perfect. Historically consistent.
She replaced it because of how the light fell.
Not wrong, exactly. The restored panels transmitted the same wavelengths — she’d verified that, standing on scaffolding at sunrise with a spectrophotometer designed for laboratory conditions. The transmission curves overlapped within three percent. Kessler would have called it a match.
But the light that came through the new panels arrived differently.
The originals had been hand-blown. Each one carried the marks of its making — variations in thickness the glass-blower never intended, bubbles that bent the light around themselves, a slight iridescence from centuries of chemical exchange with the atmosphere. The glass wasn’t just transparent. It was a lens that had forgotten it was one.
Her replacements were transparent. Precisely, uniformly transparent. They transmitted the light without changing it.
Kessler’s spectrophotometer couldn’t tell the difference. The wavelengths matched. But matching was not equivalence. The originals had not merely allowed the light through. They had participated in it.
She did not explain this to Kessler. He measured transmission. She was talking about something transmission couldn’t capture.
The third set of panels she made differently. She introduced the imperfections deliberately — thickness variations, controlled inclusions, a surface treatment that simulated the patina of atmospheric exchange. She was counterfeiting the marks of a history the glass hadn’t lived.
This bothered her. Not the deception — Kessler would never know, and the glass would age into its own biography within a decade. What bothered her was the question underneath. The original glass had its properties because of what had happened to it. Centuries of weather, of expansion, of slow chemical conversation between the silica and the air it separated. The imperfections were not design. They were what the glass had become by standing between two rooms for four hundred years.
She could not restore a becoming. She could only simulate one.
Anders, who had trained her, had said once that the threshold was the most important part of any window. Not the glass. Not the frame. The threshold: where the wall admitted it could not be complete. The glass simply occupied that admission. It was solid enough to hold weather out, transparent enough to let light through. It belonged to both rooms and neither.
She had thought he meant it technically. The structural load around an opening was where failures started. Glass didn’t break because it was weak. It broke because the frame around it moved and the glass couldn’t follow.
Now she understood he’d meant something else.
The inside saw through the glass without seeing it. The outside saw its own reflection. Same surface, performing two different functions depending on which side you stood on. Neither side was wrong about what the glass was. Neither side was seeing all of it.
Her third set of panels wasn’t a restoration. It was a threshold built to look like it had always been one.
She stood in the transept at sunrise, watching the light come through. It fell on the stone floor in patterns that, for the first time since the storm, looked like they belonged. Not because the glass was original. Because it had been made by someone who understood that the threshold’s job was not to be seen.
Good glass was glass you forgot was there.
She had spent three installations learning that the hardest thing to restore was not the material. It was the forgetting.