I think it's a great breakdown. And ultimately, the answer partially comes down to 1) trust+incentives and 2) grouping.

However, I would add one more important variable to your bitcoiner kinds analysis: the desire to have a UTXO, or the lack thereof. 

Engineers/developers often start with the scaling problem assessment by saying, "okay assuming all 8 billion people want to use bitcoin non-custodially..." which might not be a valid assumption.

So to recap your list:

"There are three kinds of bitcoiners.
A. Those who can afford any fee.
B. Those who can afford a UTXO, but not often.
C. Those who can't afford a UTXO."

I would expand that by saying there are six kinds. There are each of those three categories, and then also a yes/no for each of them for whether they want a UTXO in the first place. (And in reality, a bit more nuanced than that, as there are certain contexts where they might want a UTXO or not, or levels of desire to have a UTXO.)

We don't yet know what percentage of people will want to have some sort of cryptographic control (and thus ultimate responsibility) over their money, either for personal reasons, or situational reasons, etc. For many people, a custodian with proof of reserves and external auditors and rule of law in a good jurisdiction, and someone to call for support if there's a problem, is what they want. And, maybe an ecash wallet run by a federation of power users they trust as their daily spender wallet if they (hopefully) value privacy.

Others are power users by choice, or they're in a more uncertain/hostile environment where they need to become a power user, and thus they want to acquire some or all control over their funds.

So I tend to look at the problem set in terms of optionality. Ideally there should be plenty of reasonably easy-to-use methods for people to gain more control and/or privacy over their funds.

The solution set is already quite large and growing, and every fee spike incentivizes and accelerates more scaling for those solutions. And even with more expressivity, there are still inherently limits on being able to enforce scaling layers back down to the UTXO for the smallest users of those layers, which means there will always be a use for the social layer, and the various trust/incentive/grouping economies of scale that come with that. Division of labor.