On 23 June 2026, France recorded its hottest day ever measured, a national average of 29.8°C, past the marks of 2003 and 2019. Across the country, trains slowed and some were cancelled. The popular explanation is by now well rehearsed: steel rails expand in heat, so trains slow down to stay safe, and on saturated lines a slower train forces cancellations. All of that is true. It is also the shallow end of the story.
The interesting question is not why a train slows in the heat. It is what the slowdown reveals about how we chose to build the system in the first place. Read closely, a heat-restricted timetable is a confession. It tells you the climate band an asset was tuned for, the margin its operator decided to buy, and the bet, made decades ago, that the weather would stay roughly inside the envelope. This summer, across more and more of Europe, that bet is being called.
The neutral temperature is a wager, not a fact
Start with the number nobody outside the industry has heard of: the neutral, or stressing, temperature. Continuous welded rail, which now covers about 86% of the French network, is locked to its sleepers and ballast so it cannot freely expand (SNCF Réseau). Because it cannot move, engineers pre-set it to a “neutral” temperature, around 25°C in France, slightly lower in the Paris region, at which the rail sits free of stress. Below that, it pulls in tension. Above it, it pushes in compression, with a thrust that can reach the order of 100 tonnes before a weak point lets the track buckle sideways into an S.
Here is the part that matters for anyone who allocates capital to infrastructure. That neutral temperature is not a law of physics. It is a deliberate compromise, a single setpoint chosen to balance two opposite failure modes on the same asset. Set it high, to shrug off summer, and the rail tensions dangerously and can crack in a hard frost. Set it low, for winter, and it buckles sooner in a heatwave. France has to cover a climate running from roughly minus 15°C in winter to plus 40°C in summer, so it picks the middle. Hotter countries set their rail higher, India in the 30s, the United States higher still, precisely because they do not have to defend the cold tail. Britain stresses slightly warmer than France, and Network Rail is blunt about why it does not simply copy the desert: stress the rail for the heat and “there would be the risk of increased tension on the rails in the winter.”
So the rail is a fifty-year asset tuned to a stationary climate. Climate change does not just push the summer tail out. It widens the whole envelope and, worse, makes it move. You cannot re-stress a continental network every decade to chase it. The slowdown you see in June is the sound of an asset operating outside the band it was optimised for, and there is no cheap dial to turn.
Heat does not break the rail. It spends the slack efficiency already used
The second misread is to treat cancellations as the network failing. They are the network rationing, on purpose. And rationing bites so fast for a reason that has nothing to do with steel.
A modern high-speed railway is a triumph of high utilisation. France allocates more than six million train paths a year, packed onto cadenced lines where each “slot,” each sillon, is a precise space-time reservation and the trains run nose to tail at matched speeds. That density is the whole economic point. It is also what makes the system brittle. On a space-time graph, every train is a line with a slope equal to its speed; parallel lines never touch. Slow one train and its line tilts, leans across the slot of the train behind, and creates a conflict. On a saturated axis there is no slack to absorb it, so the only way to clear the conflict is to remove a service. Hence the familiar “one train in ten.”
This is the deep point, and it generalises far beyond railways. The efficiency that makes the network profitable on a normal day is the same efficiency that leaves no margin on an abnormal one. Heat does not introduce fragility. It reveals fragility that high utilisation had already built in, by spending, in the name of throughput, the very slack you would need in a perturbation. Tightly coupled, high-occupancy systems amplify small shocks. The 100-tonne buckling force is always latent in the steel; what the heatwave removes is the operating margin, and a system run with thin margins converts a local slowdown into a cascade. A cancelled train is not incompetence. It is triage to protect the integrity of everything still moving.
The performance paradox: the faster you run, the less you can absorb
Now the counterintuitive twist. Why does France, with some of the best infrastructure in the world, seem to suffer more visibly than slower networks? Because its strength is its exposure.
The energy carried into any track defect rises with the square of speed, the familiar one-half m v-squared. Between 20 km/h and 320 km/h the speed ratio is sixteen, so the energy ratio is sixteen squared, 256. The same two-centimetre misalignment is a harmless jolt under a shunting move and a derailment risk under a TGV. Running among the fastest trains on earth, France tolerates the smallest defects, which means it is precisely the high-speed network that must be the most cautious in heat, not the least. The performance that is the commercial advantage is also the resilience liability. This is a law of systems, not a quirk of railways: the higher you tune the spec, the thinner the margin you have left to absorb a bad day. Anyone running a high-performance operation, in logistics, in energy, in finance, should sit with that sentence.
Adaptation is a problem of tradeoffs under deep uncertainty
None of this means nothing can be done. It means every option is a tradeoff bought against an unknown future, which is a far harder conversation than “fix the trains.”
The toolkit is real and improving. SNCF Réseau has run a forecasting tool, Metigate, since 2021, predicting air and rail temperature up to fourteen days ahead across 180 sites, so restrictions can be targeted rather than blanket. Network Rail paints hot-spot rails white to shed up to 10°C, lays slab track on reinforced concrete where buckling forces are highest, and sensors the network for real-time rail temperature. Rolling stock can be renewed for better cooling. Each of these helps. None is free, and several pull against each other or against winter. The honest framing is the one hydrologists reached years ago about water systems: stationarity is dead. You can no longer design to the historical climate and assume the future resembles it. So the question that actually faces an infrastructure manager, or the public authority that funds one, is not whether to adapt but to which climate scenario to design, how much resilience to pre-purchase against a tail that is moving, and who pays for margin that, by definition, sits idle on every normal day.
The same shift, across every network
I keep noticing the same pattern from one corridor to the next. In the Red Sea, maritime risk moved from an accident you absorb after the fact to a parameter you price before you draw the route. On the rail this summer, heat is making the identical move, from a contingency you handle to a design variable you can no longer treat as exceptional. Different physics, one underlying story: infrastructure tuned to a stable world now operates in an unstable one, and the slack we spent on efficiency is exactly what we now miss.
A delayed train is a train that did not derail. But it is also a message. It tells you, in public and in real time, how much resilience a society chose to buy, and against which climate it placed its bet. The next time a TGV crawls under 40°C, the useful reaction is not to ask why the railway is so fragile. It is to ask what climate we are now designing for, and who is going to fund the gap between that answer and the network we already have.
Image by Axel-Cleris Gailloty from Pixabay