You’ve got it right. Fat cells don’t go away, except at the end of their lifespan (which I believe is around 10 years), and new fat cells can easily grow, in any case (there seems to be a genetic predispostion in some people to make new fat cells, and there is speculation that this is why some people remain metabolically healthy, despite becoming obese).
Protein and fat also stimulate insulin secretion, which in the main is a good thing, because otherwise we would die. Protein stimulates insulin secretion around half the rate at which carbohydrate does so. Nevertheless, we don’t worry about protein intake, because it is essential for life. Fat barely stimulates insulin secretion at all, in comparison to the other two macros, but it does have some effect. However, in terms of a ketogenic diet, its effect is so low that we can consider it a “safe” source of needed calories.
In the normal course of events (“normal” in the evolutionary sense, anyway), the fat cells are continually taking in energy when we eat and releasing it between meals (especially during the overnight fast). The problem with the SAD is that the constant high carbohydrate intake keeps insulin at a constantly high level, which has numerous deleterious effects on the body, and which simultaneously prevents the fat cells from releasing fatty acids to be metabolized (and prevents the muscles from metabolizing them).
The good news is that a ketogenic diet restores satiety signaling in most people to the point where it becomes very difficult, if not impossible, to overeat. There is a famous overfeeding study, focusing on high protein intake, in which one of the participants was nearly in tears at the thought of having to try to choke down one more pork chop. If I recall, correctly, he just wasn’t able to do it. Also, the men in the study gained very litttle extra weight during the study; the researchers concluded from this that the body defends its weight fairly vigorously. This is one of the reasons that the energy-balance hypothesis simply cannot be the entire story.
ETA: Forgot to add a thought stimulated by a much earlier post; namely, that protein and fat contain only approximately 4 kilocalories per gram, and fat only approximately 9 kcal/g. This, together with the inherent inaccuracy of the information in the nutritional panels of manufactured food, makes it impossible to match CI to CO with anything like the level of accuracy needed for the energy-balance hypothesiss of weight maintenance to work. And thank Heaven for that, or we’d really be [spoiler]fucked[/spoiler].