The report has the same shape: the part was fitted, it was quiet and tight, and a month later a bolt or a clamp is loose again, sometimes with a leak or a rattle that appeared gradually. Nothing was obviously wrong at fitting time. A joint that loosens over weeks is rarely one that was never tightened.
Clamp load is the force the fastener carries when tightened, and it keeps the two faces pressed together. Two things can remove it. The fastener can rotate and unwind, or the joint can settle so the same fastener now spans a slightly thinner stack. Those are different failures with different fixes, and the hardware will tell you which one you have.
Rotation happens when vibration and thermal cycling overcome friction in the threads. It is most likely on a joint with no prevailing torque feature, a reused locking nut, or a clamp whose screw is loaded in the same direction as the vibration. The giveaway is a fastener that has visibly turned.
Settling happens when the material between the faces changes thickness. A gasket compresses, a coating crushes, a burr flattens, a flange pulls slightly as it cools. The fastener has not moved, but the stack under it is thinner, so the clamp load falls away. This catches people out, because the bolt still feels tight on a spanner and the joint still leaks.
An exhaust joint runs through repeated heat cycles, and the parts either side of it rarely share the same expansion rate. An aluminium flange grows more than a steel fastener for the same rise in temperature, so clamp load is cycled up and down every time the car warms and cools. That cycling works a fastener loose where vibration alone would not.
The joint also sees the pipe move lengthwise as it heats. A clamp asked to hold two sections together while one of them slides is loaded in shear as well as tension, and the friction meant to prevent rotation is consumed by that movement instead.
Clean the head and the surface around it, then draw a line across both with a paint pen. Drive the car through a normal heat cycle and look again. A broken line means the fastener rotated. An unbroken line over a leaking or rattling joint means it held position while the stack under it collapsed, which points at settling.
Then look at the contact face. A washer or flange face showing a bright, even ring has been seated properly. One showing a partial ring, or a ring only on one side, was never square to the surface, and a fastener tightened against a cocked face stores far less clamp load than it appears to. The related hardware parts range includes the washers where these marks show up.
Crack the fastener a small amount and turn it back by hand. If it goes slack at once and spins freely, the thread is damaged or the fastener has been stretched past its elastic range. If it takes load again almost immediately and then holds, the joint is settling rather than failing. Check thread protrusion on a stud against what it was when new.
Look for galling as well. Stainless fasteners into stainless threads can seize part way in, feel tight on the spanner and clamp almost nothing. The tell is a rough, tearing feel when the fastener comes out and bright torn metal on the threads. A joint like that was never tight, so treating it as a loosening problem misses it. Thread condition is covered alongside the fabricated fittings used here.
The reliable answer is a joint that keeps its clamp load. Use a fastener with a prevailing torque feature or a properly formed locking element rather than a plain nut, do not reuse a deformed lock nut, and make sure the faces meet squarely so the load spreads. Where a gasket sits in the stack, expect it to compress.
Where the joint has to allow movement, give the movement somewhere else to happen. A slip joint with the right clearance, or a flex section close by, takes the shear out of the clamp and leaves the fastener doing the one job it is good at. That is a layout decision made before the parts are cut.
