My understanding is that the problem is that “their customer base” is not necessarily a fixed group; and that better actuarial data can assist in pruning it, or charging premiums that are an ever more accurate function of its expected costs.
In the hypothetical extreme case where the insurer has perfect information; you’d be better off just paying out of pocket and taking out a loan for anything you couldn’t afford upfront: lower overhead and no claims battling. Only if they serve a risk pooling function(either because it’s mandatory, because their information is imperfect, or both) do they provide any advantage beyond repackaging your costs into monthly payments, with overhead.
The perfect information scenario is obviously a hypothetical; but there are plenty of people who have enough information available to be identifiably lousy risks(or, in the case of pre-existing conditions, lousy known what quantities); and there is absolutely no reason to expect them to be insurable through the incentives of an insurance company.
With the less obvious risks, or the low risks, sure, no reason to kick a customer off the rolls(at least until they start making claims); but the fight has never been over “but how will we possibly get healthcare to mostly healthy people who can afford it anyway?”
