On 15 August 2026 — three years to the week after a broken wheel put a freight train on the ground inside my longest bore — the European Union Agency for Railways (ERA), the Swiss Federal Office of Transport and the wagon industry agreed to cut the freight-wheel inspection interval from 660,000 km to as little as 50,000 km, a reduction of more than 90 per cent in some cases. Switzerland applies it immediately, not from 2028; FOT director Christa Hostettler told SWI swissinfo.ch, “It is a success that we have prevailed on the substance here.”

Earn that headline back in the dark. On 10 August 2023 an SBB Cargo train — two locomotives, 30 wagons, Chiasso to Basel — lost a fragment of the right wheel disc on the 11th wagon’s first axle. Thermal overload had grown fatigue cracks inside a type BA 390 disc until it fractured; the SUST final report concluded the regime then in force could not have caught it. The dragging axle ran to the Faido crossover, struck the switch, and put sixteen of its thirty wagons on the ground.

Tape this part to the wall. I am the textbook redundant asset: 57 km, two single-track bores so one can carry the load while the other is worked, 178 cross-passages every ~325 m. Two paths — and still I lost just under thirteen months, because the derailment destroyed more than 7 km of track and an airtight safety door in the fabric the tubes share. Full service returned only on 2 September 2024; the bill was CHF 150 million. The two bores are not two railways. They share the cross-passages, the separating doors, the crossovers at Sedrun and Faido. A second path that shares a joint with the first is, at that joint, a single path.

←TODAY: Two bores, one shared safety door — and the door’s failure cost thirteen months the redundancy was built to prevent. →3012: The networks that still run carry their redundancy all the way through the joint, not just up to it. Fulcrum: Redundancy is only real where the two paths share nothing you cannot replace by Tuesday.

Now give the fix its due doubt. On 10 August 2026 the International Union of Wagon Keepers warned that the new controls touch every tread-braked wheel across roughly 500,000 European freight wagons — and that workshop capacity, not the rulebook, decides how fast they are inspected. RailMarket called it plainly: the measures need a capacity test. The cure for one shared failure mode (the wheel) is rate-limited by another shared resource (the workshops). Neither ERA nor UIP is wrong; both describe the same graph.

Atelier: The tunnel’s lesson lands on the smallest building. Two power feeds off one substation is one feed. Dual risers in one shaft is one riser. A second stair discharging into the same lobby is one stair at that lobby. This Monday, walk your live project’s redundancy diagram and mark every point where path A and path B touch the same physical thing — then ask who replaces it in year sixty, and whether the part still exists.

Hack: Price the joint. Two “independent” bores fail rarely, so true redundancy multiplies their odds toward zero — but a shared door both depend on adds its own probability straight back on top:

p_bore = 0.02          # annual failure, one bore alone
shared = 0.02          # the door/crossover both bores share
parallel = p_bore ** 2                 # "true" redundancy
effective = parallel + shared          # the joint dominates
print(round(effective / parallel))     # ~51x worse than advertised

Swap your own substation, riser or stair into shared. The multiplier shows how much of your redundancy is real.

So do the unglamorous thing: find the joint, name its owner in year sixty, cost the spare. I have carried the traffic for sections that deleted their second path to save a budget. Build the diversionary route before the wheel breaks.