Why joints fail from being too loose
Separation, cycling and the fatigue that follows.
A bolted joint fails from being too loose because an under-tightened bolt loses the favourable load sharing covered earlier in this set the moment the joint's surfaces momentarily separate under load, and from that point on the bolt takes the full external load directly and repeatedly, a far harsher cycle of stress than a properly preloaded bolt ever has to survive.
Why a loose bolt fails before a tight one
A joint with enough preload keeps its clamped surfaces pressed together throughout its working life, which means the bolt only ever sees a small share of any external load, exactly as the earlier article on load sharing described. A joint without enough preload separates momentarily every time the external load exceeds what that weaker preload can resist, and the instant the surfaces separate, the bolt stops sharing the load with anything and simply carries the entire external load alone, before the surfaces close again and the bolt's tension drops back down once the load eases off. Repeating that full swing, from a low resting tension up to the entire external load and back, on every single load cycle the joint experiences, is a dramatically harsher stress history than the small, shared fluctuation a properly preloaded bolt sees under the identical external load. The bolt in the loose joint is not simply seeing a larger swing on each cycle, it is seeing that swing anchored much closer to zero tension at its low point, which matters because a fatigue crack, once it begins to grow at all, tends to grow fastest during the portion of a cycle spent being pulled rather than compressed, exactly the portion a loose joint's bolt spends far more of its cycle inside.
The garden-gate comparison
A garden gate hung on slightly loose hinge screws swings shut with a small jarring impact every single time, the gate lifting and dropping fractionally as it swings before finally slamming against the frame, and that repeated small impact gradually works the screw holes larger with every closing, loosening the hinge further and making the next impact slightly worse than the last. The same gate hung on properly tightened hinge screws closes smoothly with no impact at all, the hinge simply pivoting under a steady, well-distributed load rather than absorbing a repeated jolt, and it can close thousands of times over years without ever loosening the way the loose-hinged gate does within a much shorter span. A bolted joint without sufficient preload is suffering exactly the loose gate's fate on a smaller, less visible scale, taking a repeated jarring full-load impact on the bolt itself with every single cycle rather than the smooth, shared loading a properly tightened joint distributes instead. A carpenter fixing the loose gate rarely reaches straight for a longer screw, since a longer screw driven into the same worn, enlarged hole often finds just as little sound wood to bite into, and the more reliable repair is usually a plug of fresh wood glued into the hole first, giving the screw genuinely solid material to grip. A bolted joint that has lost its preload through a similarly worn or damaged hole faces the identical limitation, since simply fitting a larger bolt into a hole already too worn to hold one reliably repeats the gate's own mistake rather than actually fixing it.
Why the stress range matters more than the peak load
Fatigue life is driven far more strongly by how large the swing in stress is on each cycle than by the single highest stress reached, which means a bolt cycling between a low resting tension and the full external load on every cycle can fail after a comparatively small number of cycles, while an identical bolt in a properly preloaded joint, seeing only a small fraction of that same swing on each cycle because the clamped material is absorbing most of it, can survive many times as many cycles before showing any comparable damage. This is precisely the mechanism behind a fracture surface showing the smooth, ridged marks of a slowly growing crack rather than the rough tear of a single overload, since a loose joint is not failing from any one load being too large, it is failing from the same load being repeated across a stress range the bolt was never designed to absorb that many times over.
The number that matters here
A bolt cycling through the full swing between a low resting tension and the entire external load on every cycle can reach fatigue failure after only a small fraction of the cycles an identical, properly preloaded bolt survives under the same external load, since fatigue life typically falls off steeply as the stress range on each cycle grows, rather than declining gradually in proportion to it. A joint that looks and feels adequately tight, and was simply never tightened to the preload its actual working loads require, can therefore fail in a fraction of the service life a correctly tightened, otherwise identical joint would have delivered.
What this changes in practice
Verifying that a joint has actually reached its intended preload, rather than trusting that a bolt which resists turning by hand must be tight enough, is the single most direct way to avoid this entire failure mode, since the difference between a joint that lasts indefinitely and one that fails within a modest fraction of its expected life often comes down to a preload gap invisible to feel alone. A joint discovered to have failed with the smooth, ridged marks of a slowly grown fatigue crack is worth treating as a strong sign the preload was inadequate from the start, rather than assuming the bolt itself was simply the wrong grade or size for the job, a mistake this set opened with directly, mistaking the fracture surface's real story for a material failure rather than the joint's actual clamping history.
Where this stops being true
A joint can also fail from too much preload rather than too little, a bolt tightened well past its elastic limit losing the clean, repeatable spring behaviour this whole set has relied on, and a joint deliberately designed to allow controlled separation under a known, intermittent overload, rather than never separating at all, follows a genuinely different set of rules than the ordinary case this article has covered. Preload is a target to be reached accurately, not simply a direction to keep pushing further in, and both too little and too much carry their own distinct route to failure, which is exactly the question the next set in this era takes up directly, since torque alone, the one thing most people actually control by hand at the wrench, turns out to be a surprisingly unreliable way of hitting that target.