Over the next week the plant's problems surfaced in other places: a crane that drifted when unloaded, a cutting head that fluttered at high speed, an auxiliary pump that sang at an odd pitch under heavy load. Each failure seemed small. Each nudged the same truth forward: the control architecture had been stretched thin by increased production quotas and newer, more aggressive tooling. The pressure compensators were pinned; the accumulators were undersized for the new cycle times. Systems designed for predictable loads now faced volatile demand.

Years after that, long after Peter had retired and the plant had been refitted twice over, a graduate student on a tour stopped beside the old control room. On the shelf, a battered manual lay atop a toolbox, its spine creased and its pages softened from years of reference. Someone had written one word on the inside cover in a careful hand: CALIBRATE.

"Because," he said, "it tells you what the machine will do when everything else is lying to you."

On a Sunday, while the plant hushed under dim emergency lights, a new problem arrived: the gantry motors stuttered during a rapid traverse, then recovered. Peter rode the console into the machine room and watched the scrawled plots of velocity and pressure paint a story. The integral term of a control loop was saturating and then windup was producing overshoot. He found a bypass in the feedback path: a retrofit meant to save cost had bypassed the compensator’s damping network. The machine’s response had been given a faster tempo but no dancer to hold it together.

Peter, who managed controls and liked his machines like he liked his whiskey — straightforward and no surprises — took the night shift. He walked the press like a doctor examines a patient, palms searching for heat, ears tuned to the rhythm of ancient pumps and modern valves. Nothing obvious. The PLC logs showed a spike, then a drop: a control valve hesitated.