Every few years an airline announces the world’s longest flight, collects the headlines, and then quietly kills it. Singapore Airlines did it twice. Air Canada is doing it right now with Vancouver to Singapore. Qantas is about to do the opposite and launch two of them.

The pattern is not incompetence. Ultra-long-haul routes obey a set of physical and commercial constraints that make them marginal by construction. Understanding those constraints explains every cancellation on the list, and also why the record keeps getting broken anyway.

It comes down to four things: the fuel you burn carrying fuel, the payload you cannot carry, the crew you must carry instead, and the aircraft you cannot use for anything else that day.

One: you burn fuel to carry fuel

Fuel has mass. The kerosene an aircraft will burn in hour eighteen has to be lifted, accelerated and carried for the previous seventeen hours, and lifting it costs fuel of its own. On a six-hour sector this is a rounding error. On an eighteen-hour sector it is a meaningful share of the total burn.

The consequence is that fuel burn does not scale linearly with distance. Doubling the sector length costs more than double the fuel, which is precisely why two eight-hour sectors with a stop in between are usually cheaper than one sixteen-hour nonstop. The nonstop wins on time and on what passengers will pay for it, not on cost.

Two: the payload you cannot carry

An aircraft has a maximum takeoff weight. Fuel and payload share it. Fill the tanks for an eighteen-hour sector and there is less weight left for passengers, bags and freight, which is where the revenue is.

Singapore Airlines made the trade explicit. Its standard long-haul A350-900 carries 253 passengers in three cabins. Its A350-900ULR, the aircraft that flies to New York, carries 161 in two, and has no economy cabin at all. Ninety-two seats were deleted so the aeroplane could reach the other end.

The belly is the other casualty. Freight is a significant revenue line on ordinary long-haul flying, and ultra-long-haul sectors typically leave little or no weight margin for it. The route loses a revenue stream that shorter rotations keep.

Three: you carry a second crew instead

Flight time limitations mean a sector beyond roughly sixteen hours needs an augmented crew. Four pilots instead of two, a full relief cabin crew, and certified rest facilities that occupy cabin volume which could otherwise be sold.

Those people also have to be positioned, hotelled and rested at the far end, on a route that by definition operates at most once a day. Crew cost per available seat kilometre on an ultra-long-haul rotation is structurally worse than on a route the same aircraft could fly twice.

Four: the aircraft does one thing all day

Airlines earn money on aircraft utilisation. A widebody flying two eight-hour sectors a day produces sixteen hours of revenue flying. The same airframe on an eighteen-hour nonstop produces eighteen, then sits on the ground at the far end waiting for its return slot and its crew rest.

Worse, the fleet is often bespoke. The A340-500 and the A350-900ULR exist for these missions and are awkward anywhere else. When the route stops working, the airline is left holding aircraft it cannot easily redeploy. Singapore Airlines had five A340-500s; when Newark and Los Angeles came off in 2013 the aircraft went back to Airbus.

So why do airlines keep doing it?

Because a subset of passengers will pay a very large premium to avoid a connection, and because ultra-long-haul cabins are configured almost entirely for those passengers. Singapore’s ULR is not a bad economy aircraft; it is a very good premium one that happens to have no economy at all.

The routes die when that premium demand is thinner than forecast, or when fuel moves against them, or when the geography stops making sense. Those are the three failure modes, and they explain almost every cancellation, including the one Air Canada has just announced for Vancouver to Singapore.

Qantas is betting that Project Sunrise changes the equation with an aircraft designed from the outset for the mission. It might. The A350-900ULR already proved the trick can be repeated once. What nobody has proved is that an ultra-long-haul route can survive a bad fuel cycle.

Sources: Airbus A350 and A340 programme data; Singapore Airlines fleet and cabin configurations; ch-aviation and Aviation Week reporting on the 2013 withdrawal of Singapore’s A340-500 services; Skift, “Singapore Airlines Stops Newark-Changi Flight”, November 2013.