What tightening a bolt actually does
Stretch, clamping force, and the fact that they are the same thing.
Tightening a bolt stretches it very slightly along its own length, and it is that stretch, not the friction of the threads or the pressure of the nut against the surface, which actually generates the clamping force holding a joint together, because a stretched bolt behaves like a stiff, powerful spring permanently trying to spring back to its original length.
The fastener chapter of Engineer to Win is more emphatic than anything else in the book, and it is emphatic about tightness, returning again and again to the idea that a correctly tightened bolt is a loaded spring doing real, continuous work every second it sits in a joint.
A bungee cord between two hooks
Clipping a bungee cord tightly between two hooks to hold a load in the back of a car shows the principle at a scale anyone can feel by hand. The cord's stretch is what produces the strong, steady inward pull holding the load against both hooks, and the hooks and clips only give that stretch somewhere to act. The same cord looped slackly between the same hooks holds almost nothing, even though it is still connected at both ends, because a cord that has not been stretched has nothing pulling it back toward a shorter length.
The shank as a very stiff spring
A bolt is made of metal, and metal within its normal working range behaves elastically, stretching slightly under a pulling force and springing back once the force is removed, just as a spring does, only with far less visible movement for a given force. Turning a nut down onto a bolted joint stretches the bolt's shank by a tiny, usually invisible amount, and the stretched bolt, trying to return to its natural length, pulls its head and nut toward each other with real, sustained force, as the bungee pulls its two hooks together.
That pulling force is the joint's preload, or clamping force, and it is present and working for as long as the bolt stays stretched. Picturing the bolt as a rigid pin holding two plates in line misses it entirely, since a pin with no stretch in it would contribute no ongoing clamping force at all.
A fifth of a millimetre and four tonnes
Steel is so stiff that the stretch involved is startlingly small. A correctly tightened steel bolt stretches by around two thousandths of its own length, so a bolt 100 mm long grows by about a fifth of a millimetre, roughly the thickness of two sheets of paper. That invisible stretch is enough, in an ordinary 12 mm bolt of a common grade, to produce a clamping force of around four tonnes, tens of thousands of times the weight of the bolt itself.
Because the stretch is too small to see or feel, torque is used as an indirect way of controlling it, and getting it wrong in either direction changes the clamping force the joint relies on. Too little stretch leaves the joint weakly held. Too much pushes the bolt past its elastic range, and it is the second mistake that is easier to miss, because an over-stretched bolt can still feel perfectly tight to the spanner.
Reading failures as lost stretch
Seeing a bolt as a stretched spring changes how joint failures get diagnosed. A bolt that has come loose has usually lost its stretch, whether by vibration, settling or never having been tightened enough, and a bolt that snaps under a load far below its rating has often been over-tightened into permanent deformation before it ever met the load that finally broke it. In that case the failure traces back to the moment of tightening, and the load in service merely finished the job. The bungee cord behaves the same way: one that has sagged has lost its hold on the load, and one hauled out so far that its elastic has gone baggy never pulls properly again, however firmly it is hooked on.
Past the elastic range, and soft joints that settle
The spring picture holds only while the bolt stays within its elastic range, the part of its stretch it can fully recover once the load is removed. Tighten or load it beyond that and the bolt deforms permanently, no longer springs back to its original length, and has lost some of the clamping force it was relied on to provide.
The surfaces being clamped matter as much as the bolt. A soft gasket, a painted face or any surface that compresses under load gives up a little thickness as it settles in over the first hours or days of service, and every hundredth of a millimetre it loses comes straight out of that fifth of a millimetre of bolt stretch. With so little stretch to begin with, a small amount of settling can remove a sizeable share of the clamping force the initial torque figure suggested, which is why such joints are checked and retightened once the settling has finished.