Bill Lagakos talks about that a little here: http://caloriesproper.com/?p=1912

My half-baked take on it is in the comments -- essentially that fat is a protective mechanism against an underlying disorder.

ETA: This idea is still just a fancy, but thanks to several challenges from Harry, I was motivated to find some articles that express the same idea (emphasis mine):

Abdominal Subcutaneous Adipose Tissue: A Protective Fat Depot?

In summary, whereas abdominal adiposity is associated with a higher absolute risk of metabolic and cardiovascular disease, subcutaneous abdominal fat is not associated with a linear increase in the prevalence of all risk factors among the obese. Indeed, in the case of high triglycerides, SAT may actually be a protective fat depot in obese individuals.

Abdominal Superficial Subcutaneous Fat A putative distinct protective fat subdepot in type 2 diabetes

We found that abdominal SSAT correlated with improved glycemic control and indicators of cardiovascular risk. The SSAT depot may be less lipolytic than VAT, or even DSAT, and so improved insulin sensitivity of SSAT may favor accumulation of excess energy in this depot. In this regard, higher deposition of excess calories in the SSAT is a consequence, not the cause, of improved metabolic function. Conversely, it is plausible that the abdominal SSAT fat mass may be a unique abdominal fat subdepot that has protective effects on glycemic control and cardiovascular function. This is reminiscent of a finding by some, but not all, studies that suggest peripheral SAT might be less “pathogenic” than VAT. Currently, two hypotheses have been put forward to explain the difference between peripheral SAT and VAT: The “portal theory” (6,20) implicates a direct mechanism whereby VAT is more pathogenic because its venous blood drainage is directly via the portal vein to the liver. The “ectopic fat hypothesis” (6) suggests an indirect mechanism whereby increased energy storage in peripheral SAT exerts a protective effect by decreasing fat deposition in the liver, muscle, and heart. Because abdominal DSAT exhibits an intermediate phenotype between VAT and abdominal SSAT in various functions tested (lipolysis, adipocytokine profile) (4), it is possible that these theories can underlie the unique positive association between abdominal SSAT and cardiovascular and metabolic health. Although it may be possible that peripheral SAT differs significantly from abdominal SAT, and some studies have indicated the potential pathogenic role of increased abdominal (total) SAT as opposed to peripheral SAT (9–11), further studies are required to fully understand the distinct role of the SSAT subdepot. It is tempting to speculate that discrepancies in the literature among studies assessing associations between abdominal SAT and morbidity were confined by differences in the SSAT or SSAT/DSAT distribution.

Low subcutaneous thigh fat is a risk factor for unfavourable glucose and lipid levels, independently of high abdominal fat. The Health ABC Study.

CONCLUSION: Larger subcutaneous thigh fat is independently associated with more favourable glucose (in men) and lipid levels (in both sexes) after accounting for abdominal fat depots, which are associated with unfavourable glucose and lipid levels.

Increased fat intake, impaired fat oxidation, and failure of fat cell proliferation result in ectopic fat storage, insulin resistance, and type 2 diabetes mellitus.

Abstract It is widely accepted that increasing adiposity is associated with insulin resistance and increased risk of type 2 diabetes. The predominant paradigm used to explain this link is the portal/visceral hypothesis. This hypothesis proposes that increased adiposity, particularly in the visceral depots, leads to increased free fatty acid flux and inhibition of insulin action via Randle's effect in insulin-sensitive tissues. Recent data do not entirely support this hypothesis. As such, two new paradigms have emerged that may explain the established links between adiposity and disease. (A) Three lines of evidence support the ectopic fat storage syndrome. First, failure to develop adequate adipose tissue mass in either mice or humans, also known as lipodystrophy, results in severe insulin resistance and diabetes. This is thought to be the result of ectopic storage of lipid into liver, skeletal muscle, and the pancreatic insulin-secreting beta cell. Second, most obese patients also shunt lipid into the skeletal muscle, the liver, and probably the beta cell. The importance of this finding is exemplified by several studies demonstrating that the degree of lipid infiltration into skeletal muscle and liver correlates highly with insulin resistance. Third, increased fat cell size is highly associated with insulin resistance and the development of diabetes. Increased fat cell size may represent the failure of the adipose tissue mass to expand and thus to accommodate an increased energy influx. Taken together, these three observations support the acquired lipodystrophy hypothesis as a link between adiposity and insulin resistance. (B) The endocrine paradigm developed in parallel with the ectopic fat storage syndrome hypothesis. Adipose tissue secretes a variety of endocrine hormones, such as leptin, interleukin-6, angiotensin II, adiponectin (also called ACRP30 and adipoQ), and resistin. From this viewpoint, adipose tissue plays a critical role as an endocrine gland, secreting numerous factors with potent effects on the metabolism of distant tissues. These two new paradigms provide a framework to advance our understanding of the pathophysiology of the insulin-resistance syndrome.

There is an "obesity paradox" in almost any disease, and in life itself. People in the "overweight" BMI category artless likely to die from all causes than those in the "normal" category.

Until medical science sees people's health outside of weight classes, we will continue to get these "paradoxes.". See http://www.ncbi.nlm.nih.gov/pubmed/21416445

In short: a persons cardiovascular fitness and diet are so much more important than their weight or BMI. unless the studies accounted for these factors, and if the study only looks at weight, I ignore it for being bad science.

This does not mean that getting fat is protective!

It means that if you get diabetes PRIOR to becoming overweight it's even worse for mortality risk.

Article: "In fact, it's probably not that excessive pounds are protective, said Carnethon, but rather that lean people who get diabetes are somehow predisposed to worse health."

It implies that certain people are very sensitive metabolically and must take extra care with managing diabetes risk as they can get diabetes without becoming overweight.

EDIT (supplementary):

There is a hypothesis being put forward that the ability to store fatty acids, i.e. adipogenesis, was driven by adaptations to protect against type 2 diabetes (T2D). This would imply that the frequency of T2D in ancient populations was sufficiently high to drive the selection of this function. Ignoring for a moment that other, simpler adaptations of addressing T2D are conceivable and that fat synthesis is highly conserved in species where T2D does not exist, there is no evidence that T2D was a pathology of any significance in hunter-gatherer food practices.

Unfortunately, they don't give much information about the study participants. Individuals are often diabetic for several years before they learn that they have the disease. Depending on their level of insulin production and their ability to utilize insulin, they may go into ketosis even though they are eating lots of carbs and have high blood glucose. Their bodies would be consuming fat for energy because they couldn't fully utilize the glucose. The resulting high blood glucose level would be damaging their bodies while they remained thin.

It would been helpful to know if the levels of insulin production/utilization at diagnosis was also considered in these studies.