I am as conceptually socialist and communitarian as they come but there are many reasons for the federal government not to be the identity provider.
It should set the rules (the GOVERNance) by which identity providers provide that service, but it should not itself be in that business.
My favorite way of thinking about it is- the US federal government is a singleton. In any system, you want your singletons to operationalize as little as possible, because they are hardest to change.
Another way- the US federal government is an immortal entity. It represents a perpetual accumulation of all kinds of debt- legal, administrative, technical, financial, whatever. Building and scaling new operational systems within an infrastructure consumed by debt is doomed.
The thing it can do is creating the rules and policies by which a federation of private entities can operationalize a particular need. These entities have limited lifespans, can fail, and have profit and efficiency motives, can compete for business, and are overseen and supervised.
This structure exists in lots of areas, and is more successful in some- banking- less in others- military contracting. But it's vastly preferable to that work being done in the singleton itself.
If governments are singletons, what are individuals? Maybe individuals are other objects? And now the individuals don't need to hold a reference to a government service object if they want to authenticate a message they get from another object. They just ask an identity provider object. But which one? Do individuals have a list of identity provider objects? But what if the sender is using another identity provider object they don't have? Ah! The message could contain a reference to an identity provider. But why should the receiving object trust it? Wouldn't it have to ask a government service whether the identity provider is to be trusted? No silly, we don't want a government reference! It could ask other individual objects whether they trust that identity provider object! Then cache the response? Help me out here, how does reputation work?
Seriously though, you're just moving the problem around. Adding complexity. I mean, does an identity provider object still respond to messages when it's entered bankruptcy proceedings? If you're going to use an analogy, find one that informs.
Snark aside, agree that identity is a complex problem- and compartmentalizing it into components with well-known lifecycles that have known failure modes is the right solution.
The alternative- a single monolithic identity system? No, thanks.
Note- large governmental IT systems underlying programs like Medicare and Medicaid are not operated by government employees, they are operated on a contractual basis by large IT shops. You just don't know who the operator is. That's arguably suboptimal- but a different conversation.
To the specific question- what happens in this model when an identity provider goes into bankruptcy- the same thing that happens when any entity providing critical services goes into bankruptcy.
When a consumer-facing bank fails (for instance), the bank's customers
a) don't lose their money
b) don't lose access to banking services
Their accounts are taken over by a comparable entity operating in the same geographical area.
When a critical insurance provider fails, the other entities providing comparable insurance in the operating areas have to take those contracts (even if they are terrible contracts, which they likely are, because they caused the provider to fail).
It doesn't always seem like it, but this kind of market partitioning and supervision is something that in the US both federal and most states do quite well. We should have more of it.
If you're right, then you can institute devolution of any of those things and the issues you cite are going to build up over time. It will just happen at different levels in because it's spread around some many different systems.
The unemployment issue is an excellent example of this. You want to know the reason why Congress gave a flat $600 to all UI recipients even if it would be more than they were making before? Because there isn't a single unemployment system, there are 50 systems each unique, each with their own "debt", and trying to implement appropriate strictures in all of those systems would have delayed that part of the stimulus for months, if not longer.
The way you deal with those issues is by having infrastructure that is built to deal with the issue. Call it societal/social/legal "garbage collection". Whether or not we have such infrastructure, you don't get rid of the problem by shuffling it around.
It should set the rules (the GOVERNance) by which identity providers provide that service, but it should not itself be in that business.
My favorite way of thinking about it is- the US federal government is a singleton. In any system, you want your singletons to operationalize as little as possible, because they are hardest to change.
Another way- the US federal government is an immortal entity. It represents a perpetual accumulation of all kinds of debt- legal, administrative, technical, financial, whatever. Building and scaling new operational systems within an infrastructure consumed by debt is doomed.
The thing it can do is creating the rules and policies by which a federation of private entities can operationalize a particular need. These entities have limited lifespans, can fail, and have profit and efficiency motives, can compete for business, and are overseen and supervised.
This structure exists in lots of areas, and is more successful in some- banking- less in others- military contracting. But it's vastly preferable to that work being done in the singleton itself.
Cheers.