I've spent most of my career watching infrastructure that works fine at one throughput level come apart at another. It's rarely dramatic. Nothing crashes. The system just stops being able to finish the work it was built to do, the backlog grows, and six months later somebody asks how the queue got this deep. It got this deep because the thing was sized for a volume that stopped describing reality a while ago.
Reopening structures — the formal machinery that lets you unwind a commitment after it's been made — are that kind of infrastructure. They work. They've worked for decades, resting on an assumption nobody has had to defend out loud: that reopening is the exception.
Chargebacks, for instance. A consumer gets 60 days after a statement to dispute a billing error. Card networks layer their own windows on top of that; Visa and Mastercard each define reason-specific timelines, commonly 120 days, stretching to 540 for delayed delivery. Evidence moves through intermediaries: cardholder to issuer, issuer to network, merchant responding through its acquirer. The merchant pays a fee of $20 to $100 and up whether it wins or loses. The CFPB found chargebacks nearly doubled between 2019 and 2022, from $3.2 billion to $6 billion. The structure held up fine. The cost of running it went up with every dispute.
Medical malpractice runs on a different clock entirely. Filing windows range from one to four years depending on the state, but the part that matters is when the clock starts. Most states use a discovery rule: the period begins when the patient knew or should have known about the injury. Tennessee defines accrual as the point when sufficient facts would put a reasonable person on notice. Behind that sits a hard outer boundary, a statute of repose, which extinguishes the claim regardless, usually somewhere between three and ten years. All of it rests on there being a moment when the harm becomes knowable to the person harmed.
Insurance is the most candid of the three. California's Department of Insurance tells policyholders outright that an initial property payment is often an advance rather than a final settlement, and that claims can be reopened when more damage turns up. Workers' comp allows rehearings on new evidence or a changed condition. Federal Longshore claims can be modified for a change in condition or a mistake in fact. Nobody here treats a reopening as a failure of the original decision.
What the three have in common is a set of assumptions about pace. The commitment was made at human speed. The affected person will eventually notice something went wrong. Evidence can be gathered on a timeline measured in weeks or months. And the per-case cost — the fee, the document exchange, the handoffs between intermediaries — stays tolerable because the cases are rare.
Push thousands of commitments an hour through any of that and the rare case stops being rare. Evidence gathering scales per case; so do the queues behind it. Sevda Polat made the general version of this argument in "Speed Without Contestability": review capacity degrades when decisions arrive faster than an institution can challenge them. These reopening structures are the concrete machinery that argument lands on.
The discovery assumption is the harder problem, and I don't think it has an obvious fix. Every structure above waits for a human to notice. Tennessee says it plainly — accrual begins when the facts would put a reasonable person on notice. The chargeback clock starts when the statement goes out, which assumes somebody reads the statement. The insurance supplement starts when the policyholder finds the additional damage. When an agent made the commitment on behalf of someone who may never examine the result closely enough to spot a problem, the noticing doesn't arrive late. It may not arrive. And then the only boundary left is the statute of repose, which was built as a backstop and ends up being the only clock running.