# After the Warrant: Temporal Identity, Derivative Residue, and Injurious Passage

## An independent critical supplement to the liminal-horror dossier

### 1. The proposition after independent pressure

The parent dossier makes a disciplined claim. Empty corridors, obsolete devices, decaying buildings, supernatural returns, and interface glitches remain outside its category until they govern a consequential threshold whose warrant has failed. An invitation remains legible, a route remains open or compulsory, and a system keeps classifying, recording, admitting, refusing, transporting, exposing, or commanding, although the subject can no longer reach the promised transformation, secure a terminal refusal, reverse the passage, or compel an answer. The dossier names this structure **asymmetric continuation**.

The independent research supports that proposition. It also reveals four assumptions hidden inside its first formulation.

First, recognition appears to concern whether a system can identify a subject. The new sources show a more difficult condition: the system may identify exactly the record it was designed to retrieve, but that record may describe an obsolete or imposed self. Second, residue appears to mean that a record, object, instruction, or institution persists. Machine-learning research shows that the original can disappear while its derivative influence survives. Third, passage appears to divide into admission or refusal. Housing, carceral, and disability scholarship show that admission can itself become the means of injury. Fourth, rupture appears to occur at an event: the door locks, the route loops, the guarantor dies. Slow violence, administrative burden, colonial environmental change, and residual governance show failures that accumulate, repeat, or remain deliberately unfinished.

These findings identify additional clocks inside the threshold contract. A subject changes in one time; an institution updates in another. A record is deleted at one moment; its learned derivative continues through later versions. Hardware persists according to material time; vendor support ends according to contractual time. A safety protocol retains authority from an earlier material arrangement. A colonial or archival regime declares an event past while its classifications continue to organize the present. The governing structure now includes passages whose subject, evidence, support, and authority have fallen out of time with one another.

This supplement therefore proposes a temporal extension:

> Liminal horror occurs when a consequential passage retains the power to decide after its warrant, subject model, evidentiary basis, material guarantor, or route to completion has ceased to be current—and when that temporal mismatch cannot be refused, corrected, reversed, or appealed by the person made to pass through it.

The extension remains narrow. Strong classification still requires a threshold contract, consequential passage, active residue, asymmetry, and failed completion or appeal. Stale records, statistical influence, laborious applications, and software-dependent locks enter the category only when they satisfy those tests.

### 2. Schema afterlife: recognition of the wrong temporal self

Lars Z. Mackenzie's study of trans credit reporting provides the clearest new mechanism. Credit systems use durable identifiers and long retention to establish continuity and deter fraud. For a trans person who changes a legal name or gendered presentation, that continuity can become exposure. Prior names remain attached to Social Security numbers and credit files. A mismatch can trigger suspicion in employment, housing, lending, or identity verification, subordinating present legal identity to a historical record that the system treats as evidence of a more authentic self. ([IS001](https://doi.org/10.1215/23289252-3711529))

This differs from the parent dossier's recognition failure. A facial system that cannot match a traveler commits one kind of error. A credit system that retrieves the intended file and uses its obsolete identity commits another. Its output may be administratively accurate. Its temporal relation is false. The system recognizes the wrong temporal self.

Mackenzie also identifies the cost of escape. A person may sever credit continuity and begin again, losing the favorable history that would make credit affordable. The exit from the obsolete identity can therefore require entry into predatory financial conditions. The subject faces a fork with no neutral route: preserve continuity and risk disclosure, or secure legibility at the cost of accumulated economic standing. The authority granted to continuity creates the afterlife.

Oliver Haimson and colleagues move the problem from one durable file to a network. Their study of 283 people undergoing online identity transitions describes names, photographs, gender markers, tags, and social relations that remain distributed across platform features and other users' records. Manual removal is laborious and emotionally costly. A person can edit a profile without changing a friend's photograph, a search result, an advertising category, or an automated memory. ([IS002](https://doi.org/10.1145/2858036.2858136)) The transition has no single owner because the records that represent it are relationally produced.

Platform research makes the divergence still sharper. Rena Bivens shows that expansion of user-facing gender options can coexist with deeper binary classifications, while Bivens and Haimson trace different gender schemas across profiles, advertising, and data infrastructures. ([IS005](https://doi.org/10.1177/1461444815621527); [IS006](https://doi.org/10.1177/2056305116672486)) A transition registered by the interface can therefore remain absent from the classifications used elsewhere. Os Keyes shows how automatic gender-recognition research often embeds gender as binary, immutable, and physiologically discernible before any model encounters a person. ([IS003](https://doi.org/10.1145/3274357)) The ontology determines which subjects can become valid outputs, placing the first classification failure before prediction.

**Schema afterlife** names this structure. A classification survives the identity or relation that once made it intelligible and continues to authorize decisions. **Recognition lag** names its temporal effect. Some continuity is necessary for accountability, fraud prevention, medical care, or archival integrity. The critical test asks whether the prior schema may override the current subject in a consequential passage and whether correction demands disproportionate disclosure, economic loss, or disappearance.

L. Wynholds's account of hostile trans recordkeeping sharpens the authority question. Records may be created in circumstances of mistrust, disregard for foreseeable harm, or deliberate indifference to how they will later be used. Provenance must therefore describe more than where a record came from. It must address whether the subject had agency over the uses to which that record can be put. ([IS008](https://doi.org/10.1007/s10502-025-09477-x)) A technically authentic record can be normatively hostile. In threshold terms, provenance belongs to the warrant.

The repair consequence follows directly. Adding a mutable field is insufficient if historical values retain superior authority, if other platform layers preserve another schema, or if networked records continuously return the former identity. Repair may require contextual use limits, controlled decay, propagation of corrections, protection against forced disclosure, and an appeal that lets the subject preserve legitimate continuity. The governing aim is temporal jurisdiction: deciding which representation may govern which passage now.

### 3. Derivative persistence and the erasure threshold

The parent dossier's account of digital residue focuses on profiles, records, media, infrastructure, and posthumous agents that remain after the originating relation ends. Machine learning introduces a stranger afterlife. A source datum can be deleted while a system trained from it remains.

Antonio Ginart and colleagues formalize data deletion in machine learning by comparing an updated model with one trained without the removed datum. For many standard models, they note, complete removal may require retraining from the remaining data. Their efficient techniques concern a bounded case—k-means clustering—and should not be generalized to every model. Their conceptual contribution defines deletion as the relation between the post-deletion system and a counterfactual system that did not learn from the datum, extending the terminal condition beyond removal of a database row. ([IS009](https://proceedings.neurips.cc/paper_files/paper/2019/hash/cb79f8fa58b91d3af6c9c991f63962d3-Abstract.html))

Later work demonstrates both possibility and difficulty. Chuan Guo and colleagues certify bounded removal for specified regularized linear models. ([IS011](https://proceedings.mlr.press/v119/guo20c.html)) Lucas Bourtoule and colleagues reorganize training into shards, slices, and aggregates so that later removal requires less retraining. ([IS014](https://doi.org/10.1109/SP40001.2021.00019)) Together, the results make deletion a design property whose guarantee names a model class, metric, process, and threat assumption.

Anvith Thudi and colleagues supply the darker result. Approximate or output-level unlearning definitions can be forgeable: a model artifact or training log may be made to look consistent with unlearning without establishing the claimed process. The absence of a datum's influence cannot generally be read from parameters alone. Credible assurance may therefore require algorithmic definitions and an external auditor with access to process evidence. ([IS010](https://www.usenix.org/conference/usenixsecurity22/presentation/thudi))

Two adjacent findings prevent a narrow treatment of the issue. Nicholas Carlini and colleagues extract rare training strings from a large language model, demonstrating that some ingested material can remain recoverable from a deployed model. ([IS012](https://www.usenix.org/conference/usenixsecurity21/presentation/carlini-extracting)) Michael Veale, Reuben Binns, and Jef Ausloos show a different tension: privacy-preserving designs may obstruct a controller's ability to retrieve or erase a subject's data while still failing to make adversarial reidentification impossible. ([IS013](https://doi.org/10.1093/idpl/ipy002)) One right's technical implementation can close the route needed to exercise another.

This supplement calls the resulting condition **derivative persistence**: a trained parameter, memorized sequence, inferred attribute, copied dataset, ranking, or downstream decision continues to carry force after the source record has genuinely gone. The term applies where transformation makes residue harder to locate and contest; it does not presume that every datum leaves an extractable or causally isolable trace.

The corresponding passage is the **erasure threshold**: the transition from operative influence to demonstrable non-influence under a stated standard. A deletion button represents an invitation to cross it. The platform may confirm that the request has completed. Yet the subject cannot see the counterfactual destination and may lack access to the training procedure, downstream copies, or audit evidence. When derivatives remain consequential and no one can establish or contest completion, the interface performs asymmetric continuation in an exact form. It offers an exit from the record while the record's effects keep deciding.

Different systems justify different standards: exact retraining equivalence, certified output bounds, cryptographic destruction, model-component retraining, practical approximation, or a legal prohibition on future use. Concealed equivocation creates the defect. If a service calls all of these “deletion” without naming the promised terminal condition, the subject cannot know which passage was offered.

### 4. Appeal requires an evidence route

The parent dossier concludes that repair requires access, appeal, ownership, maintenance, and governance. Independent AI-accountability scholarship specifies what an appeal has to travel through.

Inioluwa Deborah Raji and colleagues organize internal algorithmic audit as a lifecycle: scoping, mapping, artifact collection, testing, and reflection. Each stage produces a documentary trail. The authors also state the limit plainly: internal audit shares the organization's interests and can become reputation management; it is one element in a broader accountability system. ([IS015](https://doi.org/10.1145/3351095.3372873)) Timnit Gebru and colleagues' datasheets record dataset motivation, composition, collection, processing, uses, distribution, and maintenance. ([IS016](https://doi.org/10.1145/3458723)) Margaret Mitchell and colleagues' model cards record intended uses, evaluation conditions, metrics, ethical considerations, and disaggregated performance. ([IS017](https://doi.org/10.1145/3287560.3287596))

Together, these artifacts form an **evidentiary chain of appeal**. A person challenging an automated decision needs more than a locally plausible explanation of the output. The challenge must be able to reconstruct which data and schema were used, which version of the model or rule acted, what use its creators represented as valid, how it performed under relevant conditions, where it was deployed, which policy converted output into action, and who can alter the result. Without that chain, the appeal reaches a series of partial owners. The vendor points to the deployer. The deployer points to the model. The model card points to intended use. The dataset record points to provenance. No node owns the passage.

Documentation therefore becomes passage infrastructure with strict limits. A complete model card can describe a harmful use without stopping it; a datasheet can disclose a missing population without granting that population authority; an internal audit can map responsibility while leaving the responsible organization to judge itself. Sasha Costanza-Chock's design-justice account supplies the missing political condition: affected communities need material power in design alongside visibility within its paperwork. ([IS004](https://doi.org/10.7551/mitpress/12255.001.0001)) The evidentiary chain has to terminate in an actor who can correct, compensate, withdraw, or govern differently.

This distinction expands the idea of completion. Contestability requires evidence that connects the decision to an answerable authority and a possible remedy; explanation alone leaves the appeal route commemorative.

### 5. Code produces operational space

The parent dossier already concludes that code and architecture co-produce passage and defines code/space in its glossary. The independent source search adds a stronger disciplinary foundation absent from its source set: Rob Kitchin and Martin Dodge's canonical account of **code/space**.

Kitchin and Dodge argue that software and space can be mutually constituted. Software performs work that brings an operational space into being; space is continually remade through code, objects, bodies, organizations, and practice. ([IS018](https://doi.org/10.7551/mitpress/9780262042482.001.0001)) Their distinction is useful. In **coded space**, software influences a place, but the place can continue its intended function when code fails. In **code/space**, the intended spatial function depends on code. When either side of the relation fails, the operational space fails.

Martin Dodge and Kitchin's study of commercial air travel makes the proposition concrete. Travel websites, check-in, baggage, security, flight decks, air-traffic control, immigration, and customs form linked code/spaces. Bookings, identity records, policy, staff, databases, machines, and physical checkpoints repeatedly produce a traveler's route through the terminal. ([IS019](https://doi.org/10.1068/a3698)) Stephen Graham's software-sorted geographies show how ordinary systems allocate differentiated mobility, access, and service. ([IS021](https://doi.org/10.1191/0309132505ph568oa))

The crucial specification is **operational spatial failure**. A structurally sound terminal can keep its doors open, corridors lit, and flights moving while its linked system ceases to produce a valid route for one subject. The software's decision changes whether that person can enter the next room, board the aircraft, cross jurisdiction, retrieve baggage, or return.

Contemporary access-control guidance reveals why the threshold repeats. NIST's attribute-based access-control model evaluates subject, object, requested operation, and environmental attributes against policy. ([IT001](https://doi.org/10.6028/NIST.SP.800-162)) Its zero-trust architecture makes access resource-specific and continuously evaluated through a policy engine, policy administrator, and enforcement points. ([IT002](https://doi.org/10.6028/NIST.SP.800-207)) These technical forms make access a graph of recurring relations while leaving their social consequences for deployment-specific evidence.

A **recurrent authorization graph** can revoke or recompute passage after physical entry. The traveler can clear check-in and fail at security; clear security and fail at boarding; board and remain subject to border decisions at arrival. The smart-home resident can enter the house while a cloud service separately determines whether the lock, alarm, thermostat, or sensor remains functional. The threshold migrates through the route.

Code/space also supplies an essential counterweight. Staff override systems, travelers contest them, and improvisation or breakdown produces other spaces. Most code/spaces remain banal most of the time. Repeated authorization enters the liminal-horror definition only when the graph retains a consequential classification or instruction after its warrant fails and distributes responsibility so completely that no node can finish the passage or answer the appeal.

### 6. Administrative attrition and injurious admission

Threshold analysis tends to favor visible events: the gate refuses, the corridor loops, the system denies. Administrative-burden scholarship shows a passage that fails by accumulation.

Donald Moynihan, Pamela Herd, and Hope Harvey distinguish learning costs, compliance costs, and psychological costs in citizen-state interaction. Their Wisconsin Medicaid analysis demonstrates that governments can reduce or increase these burdens through policy choices and that burdens affect whether eligible people obtain benefits. ([IS022](https://doi.org/10.1093/jopart/muu009)) Herd and Moynihan's later synthesis treats administrative burden as policymaking by other means. ([IS023](https://www.russellsage.org/publications/book/administrative-burden))

The phrase matters because it displaces the fantasy of a neutral vestibule. A claimant can be eligible in law yet unable to convert eligibility into usable coverage. No single form carries the refusal. The person must learn an obscure process, assemble proof already held elsewhere by the state, repeat it after loss or transfer, absorb delay, survive stigma, and interpret silence. The process retains the appearance of an open threshold because no authoritative “no” has been issued. Passage fails as exhaustion.

**Administrative attrition** names that cumulative structure. Some burdens protect finite resources, establish identity, or enable review. The relevant test is exact: does the burden govern a consequential passage; does it accumulate asymmetrically; can it defeat a right without producing a contestable refusal; and can the institution vary it while representing the underlying benefit as available? Under those conditions, administration governs the passage itself.

Keeanga-Yamahtta Taylor's history of federally supported low-income homeownership breaks a second binary. Formal desegregation and mortgage support admitted Black buyers to a market long closed to them. Yet what Taylor calls **predatory inclusion** exposed those buyers to extractive brokers, inferior housing, unsustainable terms, repair costs, and foreclosure risk. ([IS024](https://uncpress.org/book/9781469653679/race-for-profit/)) Brenna Bhandar's study of colonial property law shows ownership regimes producing racial subjects and legal personhood through title, possession, and dispossession. ([IS025](https://doi.org/10.1215/9780822371571)) N. D. B. Connolly's history of South Florida shows how real estate and development could profit through segregation while narrating modernization. ([IS026](https://doi.org/10.7208/chicago/9780226135250.001.0001))

These histories support **injurious admission** as a specified form of the parent dossier's care/security inversion. The subject crosses the threshold, the mortgage closes, the house is occupied, and protected access arrives through an extractive relation. Admission fixes the point at which the broader parent mechanism operates.

The specification prevents a moral shortcut. Access retains material value while its allocation of asset, risk, historical classification, and exit cost remains open to analysis. Injurious admission describes that relation and does not supply a verdict against access.

Jasbir Puar's account of debility presses the concept further. State and economic power may manage populations through calibrated injury and constrained capacity rather than choosing simply between inclusion and exclusion. ([IS028](https://doi.org/10.1215/9780822372530)) Lisa Guenther's phenomenology of solitary confinement describes a death-in-life produced by deprivation of relational sense-making, while Ruth Wilson Gilmore situates prison expansion in political-economic and geographic systems. ([IS030](https://www.upress.umn.edu/9780816679591/solitary-confinement/); [IS029](https://www.ucpress.edu/books/golden-gulag/paper)) Together they disclose passages whose administratively completed destination is maintained noncompletion: the subject remains present, classified, and governed while civic or relational return recedes.

These works cannot be merged without loss. Puar's argument is politically and geographically specific, especially in its analysis of Palestine. Guenther works phenomenologically through testimony and the relational structure of being. Gilmore analyzes California's carceral political economy. A narrow synthesis survives: some thresholds govern by preserving damaged capacity, a mechanism that requires its own evidence in every case.

### 7. From gate to archipelago; from event to attrition

Alison Mountz's enforcement archipelago relocates the border. States externalize asylum through remote islands, detention centers, transit sites, carriers, processing facilities, and inaccessible jurisdictions. The right to seek asylum remains legible, but the claimant is moved away from the territory, forum, public, or institution that might complete the claim. ([IS027](https://www.upress.umn.edu/9780816697113/the-death-of-asylum/))

The threshold therefore lies in the relation among discontinuous sites. An island can function as a border within a territory; a carrier can enforce a border before departure; a database can reactivate the border after arrival. The subject keeps passing through a geography built to defer the place where an answer is owed.

This structure extends the parent dossier's airport asylum and hostile-border cases. Those cases show consequential gates and biometric passage. The enforcement archipelago shows how the gate is multiplied and displaced, moving claimants among sites whose relations keep adjudication, scrutiny, and appeal elsewhere without requiring a physical loop at any one site.

Rob Nixon's slow violence changes the time of the rupture. Environmental harm may be incremental, dispersed, and difficult to represent because it lacks a spectacular event. ([IS031](https://doi.org/10.2307/j.ctt2jbsgw)) Gabrielle Hecht's residual governance shows that waste and discarded people or places can remain under purposefully partial, inefficient administration. ([IS032](https://doi.org/10.1215/9781478027263)) Noncompletion can be governed.

Kyle Whyte places a strict limit on any universal narrative of rupture. For many Indigenous peoples, climate disruption is not a wholly unprecedented future catastrophe but an intensification of colonial environmental transformations already survived. Adaptation concerns the renewal of relationships, responsibilities, and knowledge sovereignty. ([IS033](https://doi.org/10.1215/00138282-55.1-2.153)) The threshold cannot be diagnosed by an observer who declares the apocalypse newly arrived while ignoring whose world was already broken and whose authority was displaced.

These sources support **attritional rupture**. A passage can become consequentially different through repeated burden, accumulating exposure, or the maintenance of residual status, even when no single moment bears the whole transition. The rupture criterion still requires an identifiable change in relation, a resulting effect on passage, and a demonstrated failure of completion, refusal, reversal, or appeal; duration and atmosphere cannot carry that proof.

The repair consequence is severe. Restoring function can reproduce the harm if the restored guarantor retains the authority that produced it. Whyte's emphasis on relational responsibility and knowledge sovereignty, Bhandar's account of property, and Costanza-Chock's design justice converge on one point without becoming interchangeable: repair sometimes requires transferring authority beyond service restoration. A border app that processes faster, a database that matches more accurately, or a housing program that approves more loans can intensify the existing threshold if the classification, ownership, or jurisdiction remains untouched.

### 8. The archive authors what returns

The parent dossier's mediated return describes records and images that make the absent operative again. Archival scholarship adds a prior question: who authored the conditions under which the absent could appear?

Joan Schwartz and Terry Cook argue that archives and records are active technologies of power. Creation, appraisal, preservation, description, and access shape social memory and the subjects who become legible within it. ([IS035](https://doi.org/10.1007/BF02435628)) Wynholds extends that argument to hostile provenance and the subject's agency over allowable use. Saidiya Hartman's “Venus in Two Acts” confronts a harder limit: the archive of slavery determines what can be said about lives it records through violence, and critical fabulation cannot honestly convert that limit into recovered completeness. ([IS036](https://doi.org/10.1215/-12-2-1))

**Archival authorship** names the resulting mechanism. A regime shapes residue by deciding what counts as a record, which name it bears, what description retrieves it, who may see it, and what silence will appear natural. When such a record returns through a search, verification process, memorial system, or model, it carries the institution's authorship with it.

This creates three distinct conditions that a theory of liminal horror should not merge:

1. A **recoverable record** persists and may support accountability, continuity, or memory.
2. A **hostile record** persists under a provenance or use regime that exposes or overrules its subject.
3. A **structural silence** marks what the recordkeeping regime never allowed to appear or destroyed in capture.

The first may be necessary. The second may require restriction, contextualization, correction, or subject authority. The third requires an archive to preserve the break because invented recovery would turn evidentiary absence into false closure. Hartman's refusal is decisive here.

### 9. Support withdrawal and protocol inertia

The connected home provides a material test of these temporal arguments. Julie Haney and Susanne Furman surveyed 412 smart-home users about manufacturer support ending. Unsupported devices may remain connected without security updates, clear notice, or user understanding of the resulting safety and privacy implications. Concern and intended response varied by device category. ([IS038](https://www.nist.gov/publications/consumer-perspectives-loss-support-smart-home-devices)) Within that measured scope, an installed device can outlive the manufacturer commitment on which secure function depends.

NIST's IoT capability baseline identifies software update and cybersecurity-state awareness among the capabilities required for securable devices. ([IT004](https://doi.org/10.6028/NIST.IR.8259A)) The relation is architectural when the device controls a lock, alarm, thermostat, environmental sensor, appliance, or care function. The hardware remains mounted in the home. The cloud endpoint, application, certificate, update channel, or vendor can disappear. Material life and support life diverge.

**Support withdrawal** names the remote termination of this guarantor. The stronger case occurs when support loss governs entry, exit, safety, care, privacy, or habitability; notice is absent; migration is blocked; and the object remains installed as a live dependency on an absent authority. Optional features and devices that continue safely under local control fall outside that classification.

Protocol retirement supplies a useful negative boundary. RFC 8996 deprecates TLS 1.0 and 1.1 because newer versions address weaknesses and support current cryptographic practice. ([IT003](https://www.rfc-editor.org/rfc/rfc8996)) Retirement can be legitimate maintenance. It becomes a liminal problem only when a population is compelled to use stranded clients or services, the migration route is unavailable, and the old protocol's authority continues in one part of the system while another refuses it. Obsolescence alone does not satisfy the threshold contract.

The Grenfell Tower Inquiry supplies the gravest stress test. This analysis uses its official reports solely as evidence about material guarantors, instructions, revocation, and passage, with no aesthetic classification of the fire or the deaths.

The Phase 1 report explains that stay-put strategy depended on compartmentation and reinforcing layers of protection, including a protected route. It found that continued reliance on stay-put guidance impeded evacuation after the fire escaped the conditions the strategy assumed. ([IR001](https://www.grenfelltowerinquiry.org.uk/phase-1-report)) The Phase 2 report traces the construction, product, regulatory, professional, and institutional failures that produced an external-wall system incompatible with those assumptions. ([IR002](https://www.grenfelltowerinquiry.org.uk/phase-2-report))

**Protocol inertia** describes the structural relation. An instruction rational under one material arrangement retains authority after that arrangement collapses. The system lacks an observable invalidation condition, a prepared authority to revoke the instruction, a communicable replacement, or an executable alternate route. The resident follows the authorized passage; the guarantor that made it safe no longer exists.

This mechanism adds a requirement to every safety threshold. A protocol needs more than an entry condition. It needs a revocation condition that can be observed under stress, an actor authorized to declare it, a route for communicating the change, and a practicable destination after revocation. Without those elements, care can continue as command after the basis of care has failed.

### 10. Counter-infrastructure and the limit of total systems

A theory of operative residue can easily grant too much coherence to the apparatus it criticizes. If every database remembers, every building classifies, and every protocol continues, the subject appears trapped inside a total machine. AbdouMaliq Simone's “people as infrastructure” refuses that closure. In Johannesburg, flexible relations among residents perform provisional urban functions where formal systems are fragmented or insufficient. ([IS037](https://doi.org/10.1215/08992363-16-3-407))

This supplement calls such practice **counter-infrastructure** when it reopens or substitutes a consequential passage withheld by a formal system. Reciprocal knowledge, informal routing, shared maintenance, collective witnessing, mutual care, and alternate records can produce continuity beyond the official guarantor.

The term carries an immediate danger. When residents perform functions that institutions refuse to provide, improvisation can distribute risk onto those with the fewest resources and let formal authorities treat survival as proof of adequacy. Counter-infrastructure is therefore a countercase whose evidence of non-totality must keep the unequal cost of bypass visible.

That limit changes the final account of repair. Removing the original threshold's authority, creating another route, recognizing a practice the system treated as illegible, or preserving evidence of an unrepairable break may produce a more defensible terminal condition than restoration.

### 11. Revised diagnostic

The independent research yields a stricter diagnostic for architecture and technology. A proposed case should answer all ten questions.

1. **Invitation:** What passage is represented as available, required, or protective?
2. **Transformation:** What should become different when passage succeeds?
3. **Current subject:** Which representation of the subject is allowed to govern now, and what obsolete schema can overrule it?
4. **Distributed carrier:** Which building elements, records, software, policies, services, people, and jurisdictions jointly produce the passage?
5. **Warrant:** Which authority, material condition, evidentiary basis, or support relation makes the passage legitimate?
6. **Temporal failure:** Did the warrant die, the subject change, the source record disappear, the support relation end, or the injury accumulate beyond a consequential point?
7. **Active residue:** What continues to classify, route, admit, refuse, expose, or command after that change?
8. **Terminal condition:** What observable state would count as completion, reversal, refusal, erasure, migration, or safe retirement?
9. **Evidentiary route:** Can the subject reconstruct the decision and reach an actor empowered to remedy it?
10. **Counter-infrastructure:** What practices, overrides, refusals, or alternate routes show that the apparatus is not total, and who bears their cost?

The negative boundary remains firm. Ruins, glitches, stale accounts, obsolete protocols, unsupported devices, difficult forms, discriminatory models, threatening borders, and harmful buildings qualify only when they govern passage under a still-operative residue and deny a usable terminal relation.

### Conclusion: the corridor keeps deciding

The parent dossier's central proposition survives with a more exact time structure and remains distinct from a generic theory of institutional harm.

A threshold can recognize a subject who no longer exists, preserve an influence whose source record has been deleted, admit a person through an injurious relation, distribute a border across remote geographies, dissolve operational space while leaving architecture standing, administer residue instead of resolving it, or repeat a safety instruction after its material guarantor has collapsed. Each variation retains an authority that can decide after the relation supporting its decision has changed.

The deepest new conclusion concerns evidence. Repair must extend beyond the visible point of contact: the current subject needs authority over the representation used; an erasure promise must include relevant derivatives; support and protocol need declared exit conditions; documentation must connect decisions to answerable authority; and some cases require transferred jurisdiction. Where those conditions fail, institutional time keeps the passage operative after the subject's route to completion has disappeared.

Across the cases examined here, the independent addition is a demonstrable temporal mismatch among subject, record, derivative, support relation, material guarantor, and authority, coupled to a system that retains consequential decision power and supplies no usable terminal route.