If we look at the extremes of substrate source for the provision of bulk calories we have the choice of either fat or carbohydrate.
Under high carbohydrate intake we have high pancreatic insulin output and almost matched hepatic insulin extraction. Some insulin spills over in to the systemic circulation to facilitate bulk glucose utilisation but systemic hyperinsulinaemia and hyperglycaemia should be mild and within physiological limits (whatever they might actually be...). However there is a marked differential between portal vein insulin levels and systemic insulin levels, especially post prandially.
Under extreme ketogenic conditions energy is sourced almost exclusively from lipids. Insulin has minimal involvement with hepatic glucose uptake because almost zero hepatic glucose uptake is going on. Extreme hepatic insulin resistance leads to minimal extraction of what pittance of insulin the pancreas is producing and you end up with the minimal possible difference between portal vein insulin and systemic insulin concentrations.
What happens when someone needs to use insulin to maintain normal blood glucose levels?
If your only route in for exogenous insulin is via peripheral injection you can, with ketosis, put the body in to a state where insulin is relatively unimportant. You do not have to plan for one concentration of insulin to hit adipocytes and muscles while (impossibly) targeting a far higher concentration to hit the liver. Under ketogenic conditions the liver is no longer a sump for insulin usage. In fact there is almost no sump for insulin disposal as it's not being much used for anything. Peripheral and portal insulin requirements are similar and can be met by the peripheral route.
As you move from ketogenic eating to carbohydrate based eating the portal vein to systemic insulin difference has to increase and the problems of controlling hepatic glucose output while still allowing lipolysis to give access to adipose tissue calories becomes progressively more difficult.
It's notable that successful diabetes control, as promoted by people like Dr Bernstein, uses mildly ketogenic macronutrient ratios, ultra extreme ketosis does not appear to be needed. Humans are not mice.
Carbohydrate based diets would appear to lead to that wheelchair in the dialysis room and the incorrect impression that diabetes is an inexorably progressive condition.