A desktop printer commands its motors and assumes they obeyed. There is no encoder on the axes, nothing checks that the toolhead arrived, and the controller has no way to discover that it is now printing half a millimetre away from where it believes it is. It simply continues from the coordinates in its own memory.
Everything in this category follows from that, plus the ordinary reality that a machine is an assembly of parts that loosen, wear and get hot.
The reproducibility test
Ask one question first: does the fault come back in the same place?
A defect that recurs at the same height on several different models is positional. Something physical lives at that point — a snag in the cable chain, debris on a rail, a damaged section of leadscrew, a frame member the gantry fouls. A defect that lands at random heights is load-related instead, and points at tension, grip or torque failing when the machine happens to demand the most.
That single distinction splits this category roughly in half, and it costs one observation rather than one experiment.
Fasteners, tension and squareness
Three quiet faults account for most of the rest. A pulley whose grub screw is tightened onto the round of the motor shaft rather than its machined flat will eventually spin on it. A belt with too little tension climbs its pulley teeth under hard acceleration. A gantry that is not perpendicular to the vertical axis prints every layer slightly displaced from the one below, so the part leans while every individual layer looks perfect.
None of these produce an error message. All three announce themselves as "the printer got worse", usually after the machine has been moved.
Heat is a machine fault too
Two pages here are about temperature, from opposite ends. Motors and their drivers dissipate heat continuously whether or not anything is moving, and a hot stepper loses torque exactly when a fast print needs it most. At the other end of the machine, the firmware keeps a running check on whether the hotend is behaving like something with a working heater and a working sensor attached to it.
When a printer halts with a thermal runaway or heating-failure message, that is a safety interlock doing its job. It exists because an unmonitored heater cartridge in a plastic-and-metal frame is a genuine fire risk. Guides that suggest disabling the check, widening its tolerances or commenting it out of the firmware are telling you to remove the smoke detector because it keeps going off. Find the loose thermistor or the failing cartridge instead.
Wear is not a defect
Belts round off their teeth. Brass nozzles open up, especially after any filled filament. Bed springs relax. Sealing faces get scored by repeated cold tightening. These parts are consumables on a machine that runs regularly, and a printer that has become gradually less accurate over a year of daily use may need replacements rather than recalibration.
Knowing when to stop tuning
The most expensive failure mode in this category is not the fault — it is spending a fortnight on slicer settings for something no slicer setting can reach. If a defect survives a profile you know is good, appears on models sliced by different software, and is repeatable in position, stop opening settings and go and move the gantry by hand with the motors disabled. What you feel and hear in those thirty seconds usually ends the investigation.