Everything in this category comes down to one competition: the strength of the bond between the first layer and the plate, against the accumulated contraction force of everything printed on top of it.
The bond is roughly fixed once the first layer is down. The force is not — it grows as the part grows. That is why some prints fail immediately and others run beautifully for two hours and then let go, and it is why "it stuck fine at the start" is not evidence that adhesion was adequate.
The two halves of the problem
Getting the first layer down. A first layer bonds by being pressed into the plate's surface texture while still soft. Three things stop that happening: a nozzle too far from the plate, so the extrusion never gets squashed; a contaminated plate, so the plastic touches grease rather than the surface; and a bed too cold for the material, so the plastic solidifies before it has keyed in.
Skin oil is the most common contaminant and the least appreciated, because isopropyl alcohol does not remove it. Dish soap and hot water do.
Keeping it down. Every layer laid on top contracts as it cools, pulling on a base that has already finished contracting. On a low-shrinkage material like PLA that force stays small. On ABS, ASA, polycarbonate or nylon it becomes large enough to peel a corner off a properly bonded plate, and the answer is to reduce the force — an enclosure, a hotter bed, less part cooling — rather than to keep increasing the bond.
What is in this category
Warping and print detachment sit here because they are adhesion failures even though they look like material problems. So do the first-layer-too-high and first-layer-too-squished pages, which are opposite errors with the same cause, and the two "the adhesive did not help" pages, because an adhesive applied to a cold plate or applied too thickly makes things worse rather than better.
Spaghetti failures are here too, though they are always a consequence rather than a cause: the part came loose or a layer shifted, and the printer kept extruding into air. Diagnosing the spaghetti is pointless; diagnosing what let go before it is the useful work.
The order worth working in
Clean the plate. Check the Z offset with a single-layer test patch rather than on a real part. Get the plate hot enough for the material, and do not let a profile cool it partway through the job. Then, and only then, add a brim.
A brim is a mitigation for a marginal bond, not a substitute for a good one, and a brim on a dirty plate lifts with the part.
Why this category is the cheapest to experiment on
Every fault here declares itself within the first few minutes of a print, and every one of them can be reproduced by a single-layer patch that takes about a minute of filament. Nothing else on this site gives feedback that fast.
So the habit that pays off most is to stop testing adhesion changes on real parts. Change one thing, print the patch, look at it, and repeat. An afternoon of that will teach you more about your own plate than a season of failed overnight prints.