Let me reinforce what Ed said about varying cross sections as a major contributor to distortion. On heating steel expands until you reach the point where things start going into solution, or what most knifemakers call “critical”, at this point the steel noticeably contracts until the conversion is complete. If there are thin sections attached to thicker sections obviously one of those sections will be contracting while the other is expanding if the heating is fast and continuous. This is why industry employs things like pre-soak holds to equalize the temperature throughout before proceeding with the high heat. On cooling the steel is contracting until you reach the temperature at which it begins to harden (in this case around 400F) at which point it expands drastically and rapidly, like speed of sound rapidly. Here the same is true as before of differing cross sections, thin and thick are going to move in opposite directions. This is why industry employs things like marquenching to keep complex shaped parts useable. Since I use the marquenching process I just did a batch of blades that included two kitchen knives that had edges ground to .010”. By quenching to 420F and allowing the whole cross section to equalize before continuing the hardening process they came out straighter than many other blades I do, and with a lovely 66HRC. In a conventional quench the edge would be expected to take the all too familiar ribbon effect.
These are just a couple of the most common contributing factors to distortion, but the topic is complex and endless with any number of factors, including the phase of the darn moon for all we know. In that same batch of knives I had a longer hunter blade come out of the quench and before my very eyes warp more severely than any knife I have ever made. I started to straighten it but watched more “s” curves begin as I worked it, so instead I put it in the vice and intentionally snapped it. With that amount of unexplainable distortion it was obvious that that knife had serious variables that I could not account for and thus I had no idea how badly something could be off inside it. The fractured grain size was as smooth as glass and the homogeneity was impeccable for damascus, but obviously there was something going on there. But the snapping itself gave me some data on the strength to ductility curve during the martensitic transformation, something that is so brief that we rarely get to explore; if I had waited another 8 minutes the break would have occurred in an entirely different manner.
The fact is that you could do everything right and still have some distortion. Steel is not a perfect substance and is made up of inconsistent structures. The best way to avoid it is to do everything as evenly as possible, heating, cooling, grinding normalizing, annealing, stress relieving etc… to keep the strain energies stored in that inconsistent structure balanced the best we can.