Going up is a matter of power. Coming down is a matter of life.
Everything a train gains in height on the way up it must lose on the way down, and the energy has to go somewhere. On a rack railway it cannot be left to the wheels: they would simply slide. So the engine holds the train back through its pinions, and turns the energy into heat.
Riggenbach's engines used counter-pressure: the driver puts the engine into reverse while it rolls forward, so the cylinders pump air against the pistons and act as a brake, with water sprayed in to keep them cool. Band brakes on drums beside the pinions hold the rest, and an automatic brake fires if the train runs too fast.
Keep under 5 mph. Over 7 mph the automatic brake fires. Counter-pressure only works while the wheels turn, and grows with speed; the hand brake also holds the train still, but heats up and fades if you lean on it. Shown ten times faster than real.
Drop 960 metres with 37 tonnes and around 350 megajoules must be turned into heat on the way down: enough to boil a one-litre kettle a thousand times.
The Snowdon Mountain Railway opened in the spring of 1896 with three Swiss engines on the Abt rack, and Engineering had stressed that "no appliance has been spared" for safety. On the very first day for paying passengers, it went wrong.
At the inquest the jury could not say why the engine mounted the rack. Sir Douglas Fox blamed the settlement of the track after the thaw. A passenger wrote to Engineering doubting that the rack teeth had been laid to the drawing; another blamed the brakes. Abt's firm wrote to defend the system, pointing out that the same rack and engines had run without trouble on Monte Generoso, Rochers-de-Naye, the Rothorn, the Schafberg and Pikes Peak.
The carriages survived because they were not coupled to the engine.
Look back at the meshing lab in chapter 4: a quarter of every push lifts the pinion. Coming downhill the pinions push hardest of all, holding the whole train back, and every millimetre the track drops away takes a millimetre off the 36 the teeth have to work with. The gripper rails do not stop that lift happening; they simply make sure it can never be enough.
Sources: Engineering, vol. 61 (1896): the line description, the leader of 17 April 1896 "The Snowdon Railway Accident" and letters of 13, 20, 27 and 28 April. The accident's date is from a letter of 13 April that says "Monday last". The reopening in 1897 and the gripper rails come from later histories of the line, not from Engineering. V. von Röll (1923), "Zahnbahnen", for counter-pressure braking.