Every fault in this category is the same disagreement: the volume of plastic the machine was told to deliver, and the volume that actually left the nozzle.
The reason that disagreement is possible at all is that nothing measures the output. The slicer converts a path into a volume, the firmware converts that volume into a rotation of the drive gear, and the drive gear turns whether or not any plastic moves. There is no sensor at the tip. An extruder is a pump running open loop, and every page below is a way for that assumption to break.
Which one you have is usually answered by when, not by what it looks like.
A constant error means calibration
If every print is wrong by roughly the same proportion — walls always a little fat, or always gappy — nothing is failing. Something in the chain of assumptions is simply wrong: the diameter entered in the slicer, the extruder's steps per millimetre, a flow multiplier copied from another machine, or a nozzle that has been quietly worn larger by abrasive filament.
Constant errors are the easy ones. They are measurable with calipers and a single-wall test print, and once corrected they stay corrected.
A progressive error means something is heating up or filling up
A print that starts perfectly and degrades is the most informative case on this page, because the clock itself is diagnostic. Failure after twenty minutes to an hour, at no particular height, points at heat migrating up into a part of the filament path that is supposed to stay cool. Failure that creeps in over many hours points at material slowly carbonising in the melt zone or an abrasive gradually opening the orifice.
Note the shape of the evidence: not what the print looks like, but how long it survived. That is worth writing down when a print fails, because it is the piece of information people forget by the following morning.
An intermittent error means something outside the hotend
Flow that comes and goes in bursts — audible ticks, patchy walls, brief gaps that recover on their own — is rarely the hotend at all. Water flashing to steam inside the melt zone does this. So does a spool that binds and then releases, filament with poor diameter tolerance, and a drive gear that has already chewed a flat into the filament and grips only intermittently.
Grinding and clicking are consequences, not causes
Two of the pages here describe an extruder that is failing to move filament: a chewed flat on the filament, and a rhythmic click from the motor. Both are the drive system reporting that it has run out of force, and the resistance in front of it is what needs diagnosing. Tightening the idler harder is the intuitive response and it makes both worse — it deepens the flat and it deforms the filament without addressing whatever is blocking the path.
Every hotend has a ceiling
Beyond a certain volume per second, melting is the limit rather than motor torque. That ceiling is a property of the hotend's geometry and its heater, and it is why identical settings behave differently on two machines. The print time estimator reports the flow a given combination of speed, layer height and line width actually demands, which is the number to compare against the machine's real capability before blaming the filament.
Where to start
Weigh the evidence in this order: is it constant, progressive, or intermittent? Then check the free things first — the spool turns freely, the drive gear is clean, the filament is dry. A remarkable proportion of the flow faults people bring to forums are solved by one of those three before any setting is opened.