There's some challenges with underwriting and asymmetrical payoffs that are magnified on the exchanges since, my understanding (which may be off) the flexibility with rates is much more limited on the exchanges.
To make an analogy using a less polarizing insurance, home owners insurance in a very simple world. The biggest risk with home insurance is distance to coast since hurricanes are a covered risk. Proximity to coast would provide your FREQUENCY or LIKELIHOOD of a major loss. The other risk is the value or size of your home, this informs the SEVERITY of a potential loss. So let's consider what that does to insurance. Mansion on the coast - high risk so I charge high premium. Two bedroom ranch in the middle of the desert - low risk so I charge a low premium. So now we know risk and premium are correlated.
Now lets consider the asymmetrical payoff. Let's say I charge you $1,000 to insure your $10M home (exaggerated for illustration) and your deductible is $5,000. In any given year your maximum loss is $6,000 (premium + deductible) so if your entire house is obliterated in a hurricane, I am on the hook for $10M. You essentially paid $6,000 to "make" $10M (emotional toll and inconvenience aside).
In health insurance exchanges there are limitations to underwriting, partially due to incomplete data (e.g., I may not know your most recent blood panel) and partially due to restrictions (IIRC, the highest and cheapest insurance have to be within a factor of 3). So there are situations where the lowest risk policies are over-priced but the highest risk policies are under-priced. Curve ball number one: the cost of the policy for the lowest risk population is too high relative to the penalty, so there's little to no incentive to buy insurance "for nothing". As a result there is a selection bias towards the higher risk, under-priced policies. Curveball number two: even the stingiest health plan has an out of pocket maximum which caps the policyholder's exposure to financial loss in the event of a severe episode, which is more likely for these higher risk populations. Need a $500K liver transplant? That'll be $5,000 dollars please. Had a $250K NICU baby? That'll be $15,000 please. Are you a hemophiliac and have a $300K+ Rx tab every year? $5,000 for you too. This is that asymmetry again.
Lets combine the two: over-representation of under-priced, high risk policyholders and they have stop-losses but the insurance company doesn't. It doesn't take much for this to tank the loss ratio(s) for the product(s) and if it tanks hard enough, it can drag down the loss ratio for the insurance company as a whole.
Without getting into politics or other dimensions of this complex topic, it seems obvious why a for profit company would pull out of this, especially if we use home insurance to ground the conversation.