Distributive Capture

A threshold concept for infrastructure dependence and distributive control

ContentsBrowse sections

Basis of analysis

Conceptual claims rest on the infrastructure, capture, gatekeeping, path-dependence, and transaction-cost literatures listed in the References, which define neighboring concepts and supply rival explanations while carrying no empirical verdict about the default Ethereum access path. Product configuration rests on living MetaMask and Infura documentation retrieved on July 15, 2026; that packet records retrieval metadata, reported response fingerprints, researcher-selected extracts, and any access failure separately from the live pages, and it omits the response bodies, so a reviewer cannot recompute the fingerprints or check the extracts against the absent pages. One dated instrument reading rests on the May 17, 2026 repository packet, which preserves a raw Relayscan JSON response, capture metadata, cryptographic hashes, and a deterministic replay check; that custody fixes which bytes were examined and when, and the reading itself is stale, preserved for reproducibility rather than current market description. The Ethernodes routes captured in that session returned Cloudflare challenge pages, while Rated's unauthenticated API returned an authorization boundary and its explorer route returned a checkpoint, so no dashboard value from either source enters the analysis. Printing, rail, and internet scholarship enters as genealogy and comparative pressure, never as a pooled causal demonstration. The assembled record holds no original on-chain collection, no wallet telemetry, no measured substitution comparison, no rule-bound discipline-event corpus, and no code checkout or deployment observation, so the findings reported below stand as later-state components covering residual mediation, configured dependence, and governance relevance, while Admission, effective bypass, and effective discipline remain unresolved and capture remains unestablished.

Abstract

Distributed systems often widen access through infrastructure that remains outside their formal core. Wallets, hosted nodes, interfaces, relays, custodians, and coordination venues make protocol participation practical. They also create positions from which access, ordering, continuity, or accountability may be conditioned. Distributive capture names a proposed threshold within that sequence: a residual carrier becomes decision-bearing over participation, while the affected participant class lacks an effective route of bypass or discipline.

The term is narrower than regulatory capture, obligatory-passage-point analysis, gatekeeping, infrastructural power, path dependence, and transaction-cost explanation. Each adjacent literature identifies part of the problem. None, by itself, supplies the conjunctive threshold proposed here. Concentration, default status, and even an observed act of exclusion can establish governance relevance without establishing capture.

The article applies the protocol to the default MetaMask and Infura access path. Current official materials document a default provider, specified regional restrictions, and configurable alternatives. Those materials support residual mediation, configured dependence, and governance relevance as component findings. They do not establish a prior bottleneck and observed widening for a specified participant class, measure substitution, or supply a complete discipline-event corpus. Under the ordered protocol, the case stops at unresolved Admission.

No case examined here clears the threshold. Distributive capture remains a diagnostic protocol with an empty confirmed extension. A dated builder-market snapshot is retained as an instrument reading, with its custody and inferential limits stated separately. The former Bitcoin negative case and the proposed Solana/Firedancer prediction are withdrawn until matched comparative evidence and a dated adoption baseline exist.

Keywords: distributive capture, residual infrastructure, regulatory capture, obligatory passage points, gatekeeping, infrastructural power, Ethereum, MetaMask, Infura, research design.

1. Claim and Evidentiary Stop

Permissionless protocols do not eliminate mediation. They relocate it. An ordinary user may hold protocol-valid credentials and still reach the system through a wallet’s default RPC connection, a hosted interface, a block builder, a staking service, or another operational carrier. These services can widen practical access. Their importance therefore cannot be inferred from formal protocol authority alone.

Infrastructure scholarship has long treated infrastructure as relational: it becomes visible through use, dependence, and breakdown rather than through a component list (Star and Ruhleder 1996; Edwards 2003). Hughes (1983) explains how technical systems accumulate organizational momentum. Jasanoff’s (2004) idiom of co-production directs attention to the reciprocal formation of technical order and social authority. These works motivate the inquiry. They do not establish that any contemporary intermediary has crossed a capture threshold.

The narrower object is residual infrastructure conversion. A function remains residual when ordinary participation depends on it even though the protocol does not fully specify, provide, or govern it. Conversion occurs when the carrier of that function acquires practical authority over a term of participation. The relevant terms are access, ordering, continuity, valuation, legitimacy, and accountability. Appropriation can intensify the case, though profitability or rent extraction cannot substitute for evidence of participant dependence.

Distributive capture is the proposed terminal state. It requires a satisfied Admission burden (a defined capacity, prior bottleneck, specified participant class, and observed widening), followed by an operationally necessary residual function, participant-path dependence, an observed governance-relevant capacity, and failed bypass and discipline for that same class. The sequence is cumulative; a later component cannot cure an earlier gap, and the last two burdens are conjunctive.

This formulation changes the status of the Ethereum illustration. Official product materials show that the default MetaMask path uses Infura, can transmit provider or jurisdictional restrictions before a request reaches Ethereum, and offers configurable alternatives. They do not establish the Admission comparison, a controlled test of those alternatives, or a complete record of attempts to contest the relevant terms. The path supplies governance-relevant component evidence, but the ordered case remains unresolved at Admission.

The title therefore names a research hypothesis. It does not announce a finding about Ethereum, blockchains as a class, or digital infrastructure generally. An empty confirmed extension leaves the concept vulnerable to redundancy, elasticity, and empirical failure. The protocol below makes those vulnerabilities inspectable.

2. Evidence Architecture

The unit of analysis is a participant path: the sequence of operational services through which a defined participant class performs a defined task. A wallet default can identify a designed path. It cannot establish population prevalence without usage evidence. A dashboard can describe an observed measurement window. It cannot establish persistent authority without a temporal series and a mechanism linking the measured share to decision-bearing conduct.

The present study does not include original on-chain collection. It combines conceptual analysis, current official product documentation, and dated source captures assembled for earlier evidence packets.

The evidence is ordered in three layers.

First, conceptual claims are grounded in the relevant literatures. These sources define neighboring concepts and furnish alternative explanations. They carry no empirical verdict about the default Ethereum access path.

Second, living official materials establish product configuration and documented alternatives. The July 15, 2026 refresh records researcher-selected extracts, retrieval metadata, reported response fingerprints, and any access failure separately from the live pages. Because it omits the response bodies, the packet is an observation log rather than independently auditable claim custody: a reviewer cannot recompute the fingerprints or verify the extracts against the absent pages. It also cannot supply a publication year, prove user prevalence, measure reconfiguration difficulty, or establish provider competition.

Third, repository evidence preserves dated instrument outputs. The May 17, 2026 packet contains a raw Relayscan JSON response, capture metadata, cryptographic hashes, and a deterministic replay check. Source custody establishes which bytes were examined and when. Empirical proof requires additional warrants: construct validity, coverage, error analysis, persistence, causal relevance, and a justified inference from measurement to governance.

No claim in this version depends on a current Ethereum, Bitcoin, or Solana code checkout. The article makes no inference from repository code to deployed software, validator adoption, wallet configuration, or network behavior. A source clone would record a commit state; it would not prove deployment. For that reason, no clone boundary is presented as empirical evidence. Any future code-dependent claim must record the repository URL, exact commit, submodule state, dirty status, build configuration, and a separate deployment observation.

Historical material occupies a fourth and limited role, supplying genealogy and comparative pressure, never a pooled causal demonstration. Printing, rail, and internet scholarship can show that expanded circulation has repeatedly depended on institutions that organize access. It cannot establish a single invariant mechanism across those domains.

The result is not a demonstrated causal sequence across every historical period discussed or a fully closed blockchain-class proof. Every contemporary claim remains bounded to its participant path.

3. What the Concept Adds

3.1 Regulatory Capture

Regulatory capture begins with a public regulatory institution. Stigler (1971) modeled regulation as something industries seek and use for their benefit. Peltzman (1976) generalized the political exchange among organized interests. Carpenter and Moss (2014) place the modern concept around the displacement of regulation from a public purpose toward the interests and intentions of the regulated industry.

Distributive capture begins elsewhere. Its candidate carrier is an operational intermediary within a distribution-enabling system. No delegated public mandate is required. The alleged authority emerges from control of a residual participant path. The object being captured is practical participation, not a regulator.

The distinction yields a clean test. If a service provider influences an agency that regulates it, regulatory-capture analysis applies. If the provider’s infrastructural position lets it condition user access, the distributive protocol may apply. Both processes can occur in the same setting, yet evidence for one does not prove the other. A sanctions-based regional restriction, for example, may show that legal authority can be transmitted through a provider path. It does not show that the provider captured the state or that the user path has crossed the distributive threshold.

The word capture therefore carries no presumption of corruption, collusion, or unlawful conduct in this account. It describes a structural outcome that still requires proof of failed bypass and failed discipline. That choice also creates a burden: if the protocol cannot distinguish structural capture from ordinary service provision, the term should be abandoned.

3.2 Obligatory Passage Points and Gatekeeping

Callon’s (1986) obligatory passage point identifies a relation through which actors must pass for a network project to hold together. It is a theory of translation and positioning. An obligatory passage point can locate where leverage may form. It does not determine whether alternatives are practically unavailable, whether the position is durable, or whether affected actors can discipline its holder.

Network gatekeeping concerns control over passage, selection, and information flow. Barzilai-Nahon (2008) extends gatekeeping to networked environments in which gatekeepers, gated actors, and their relations vary across contexts. A provider’s refusal, filtering decision, or routing rule can be a gatekeeping act. Within the distributive protocol, such an act is evidence for governance relevance. Capture still requires the additional participant-class tests.

The concepts therefore operate at different analytical moments:

an obligatory passage point identifies a strategically central relation;
gatekeeping identifies a selection or exclusion capacity exercised at that relation;
distributive capture alleges that a distribution-enabling residual relation has hardened beyond effective bypass and discipline.

This sequencing avoids two errors. Centrality does not equal control, and an observed gatekeeping act does not establish durable capture. The relevant carrier must also be fixed at the correct level. An Infura-specific claim, a hosted-RPC-market claim, and an RPC-function claim have different alternatives and different evidence burdens.

3.3 Infrastructural Power, Path Dependence, and Transaction Costs

Mann’s (1984) infrastructural power concerns the state’s capacity to penetrate society and implement decisions through logistical organization. Distributive capture tracks authority running the other way, from an operational carrier toward a participant path. A state restriction transmitted through a provider may involve both directions. Legal command belongs to the state, while path-specific implementation reveals governance relevance at the provider layer.

Pierson (2000) explains how increasing returns, sequencing, and institutional feedback can stabilize a path. Path dependence can explain why switching becomes difficult. It does not specify when difficulty becomes decision-bearing authority, and it does not identify the actor or layer that must be disciplined.

Williamson’s (1985) transaction-cost account supplies an important rival explanation. Hosted infrastructure may dominate because specialization economizes on scale, uncertainty, and asset specificity. Such an arrangement may remain competitive and contestable. Evidence of efficiency, concentration, or high self-provisioning cost cannot establish capture until provider substitution, market entry, collective voice, and re-internalization have been examined.

Hirschman’s (1970) distinction between exit and voice clarifies the two terminal burdens. Exit concerns practical bypass; voice concerns attempts to change the incumbent term. Olson (1965) explains why a diffuse participant class may struggle to organize voice even when its aggregate stake is large. The protocol uses these mechanisms explicitly. It does not treat silence as evidence of failed discipline when the archive lacks an event universe and an observation rule.

The wider decentralization literature supplies the substantive problem. Musiani (2015) shows how service operation can centralize governance around decentralized architectures. De Filippi (2020) separates technical decentralization from decentralization of power. Ferreira, Li, and Nikolowa (2023) analyze corporate capture within blockchain governance. Distributive capture adds value only if its residual-function sequence and conjunctive threshold discriminate cases these broader accounts group together.

4. Diagnostic Protocol

The protocol has five ordered states. Each state adds an empirical burden. The case cannot advance past its earliest unresolved state, although later component evidence may be reported separately so long as it is not converted into an ordered verdict.

State	Question	Minimum evidence
Admission	Did a technical-institutional system widen practical access to a previously bottlenecked capacity?	A defined capacity, prior bottleneck, participant class, and observed widening. Promotional language alone is insufficient.
Residual mediation	Does the participant’s task traverse a function outside the formal core?	A path reconstruction identifying the function, carrier, task, and protocol boundary.
Dependence	Does the specified class ordinarily rely on that path?	Usage evidence, a documented default with its limited meaning stated, or another justified indicator of reliance.
Governance relevance	Can the carrier shape a term of participation?	An observed or directly documented capacity affecting access, ordering, continuity, valuation, legitimacy, or accountability.
Capture	Have effective bypass and effective discipline both failed for that class and task?	A preregistered comparator study plus a complete, rule-bound event record. The carrier and level of analysis must remain constant.

Capture is a conjunctive verdict. It requires all prior states, failed bypass, and failed discipline. A path with unknown bypassability remains unadjudicated. A path with no qualifying provocation has an unscored discipline arm. No number of concentration snapshots can fill either gap.

A governing rule follows: the threshold into capture must be argued path by path. Evidence cannot migrate across participant classes, tasks, carriers, or layers without a separate warrant.

4.1 Bypass

Bypass is practical substitution for the same task and participant class. The relevant alternatives depend on the alleged carrier.

For an Infura-specific claim, another hosted provider is a valid candidate exit. For a hosted-RPC-market claim, provider switching stays within the alleged layer, so self-hosting or another architecture becomes relevant. For an RPC-function claim, the inquiry is broader again. Collapsing these levels lets the claim survive every counterexample by changing its object after the evidence arrives.

The comparison must be registered before outcomes are inspected. It should specify:

participant class and technical competence;
task and required security properties;
incumbent path and named alternatives;
setup time, recurring labor, direct cost, reliability, latency, error rate, privacy, and feature loss;
measurement window and workload;
minimum acceptable service floors;
missing-data treatment and decision rule.

Universal thresholds such as thirty days, a fixed cost premium, or a fixed percentage of service degradation have no warrant in the current record. Appropriate thresholds must derive from the task, the participant population, observed performance distributions, and a declared loss function. A consumer payment task, a small application’s production endpoint, and an institutional validator operation have different feasibility boundaries.

4.2 Discipline

Discipline concerns the capacity to alter the incumbent term without abandoning the distributed capacity. Qualifying mechanisms can include organized voice, provider competition, wallet-default change, audit, governance intervention, re-internalization, or a technically credible fork. The evidentiary object is a defined set of provocations and responses.

An event study must state its inclusion rule. Examples include every documented service restriction affecting the participant class during a fixed period, every provider outage above a declared duration, or every material routing-policy change announced through named channels. For each event, the record should track user switching, public contestation, provider response, default changes, entry by rivals, and persistence of the term.

Two salient episodes with no observed reversal cannot establish failed discipline when the search process is undefined. Absence of a recorded response may reflect archive incompleteness, an event too small to organize against, successful provider substitution, or participant indifference. The discipline arm is scored only when the event universe and observation window make those interpretations testable.

4.3 Interpretation Rules

Three rules prevent threshold drift.

First, the carrier stays fixed. Evidence about one company cannot establish capture by an industry, and evidence about an industry cannot establish capture by a technical function.

Second, governance relevance stays distinct from motive. A path can transmit a legal command, implement an operator policy, fail technically, or reflect a contractual limit. Each event may condition participation. Their causal interpretations differ.

Third, every unresolved burden remains visible. Unknown evidence is never scored as failure. Later-state component findings retain their descriptive value, but the ordered classification ends at the earliest unresolved state.

5. Worked Application: MetaMask’s Default Ethereum Path

The adjudicated unit is narrow: a MetaMask user attempting an Ethereum task through the wallet’s configured default provider. It is not every MetaMask user, every wallet, every RPC provider, or Ethereum as a system.

5.1 Residual Mediation and Dependence

MetaMask Support states that MetaMask accesses the blockchain through Infura by default and describes Infura as the default provider (MetaMask Support n.d.-a; n.d.-b). Infura’s own FAQ states the same configuration and identifies switching to another provider or a user’s own node as available options (Infura n.d.). Together these sources establish a designed default path.

RPC provision is residual to Ethereum consensus. The provider receives and returns requests that let the wallet read network state and submit transactions, while consensus rules remain elsewhere. The official materials therefore support residual mediation for the specified path.

The dependence claim is limited to configured dependence: absent reconfiguration, the default structures the initial path. The archive contains no wallet telemetry showing who remains on that path, for which tasks, or for how long. Nor does a configured default establish Piersonian path dependence, which would require increasing returns, sequencing, or feedback that stabilizes the path over time.

5.2 Governance Relevance

MetaMask Support states that Infura is unavailable to persons in certain jurisdictions because of comprehensive United States sanctions programs (MetaMask Support n.d.-a). Its connectivity guidance also acknowledges intermittent default-provider connectivity and advises users to wait, connect to another provider, or use another client (MetaMask Support n.d.-b).

These materials establish that a provider-layer condition can determine whether the default path reaches Ethereum at a particular place or time. That is governance relevance in the access and continuity predicates. The finding remains path-specific. It does not imply a change to Ethereum consensus, a provider monopoly, corporate authorship of the legal restriction, or regulatory capture.

The source also reveals why motive and capacity must be separated. A sanctions-based restriction anchors in law outside the provider and travels to the participant through infrastructure. Its existence shows where an authoritative term attaches. It supplies no evidence that the provider sought the term or benefits from it.

5.3 Bypass Burden

The official record documents three alternatives: another provider, a local node, or another client. MetaMask permits multiple RPC URLs and a changed default (MetaMask Support n.d.-c); a custom RPC can disable Smart Transaction features such as MEV protection, while local operation requires execution and consensus clients (MetaMask Support n.d.-d). This establishes technical alternatives and possible friction, not effective bypass. No participant experiment compares completion, cost, failure, feature loss, security, privacy, or sustained reliability.

Carrier definition controls the inference. Provider switching is exit from Infura; a hosted-RPC-market claim would instead require competition and self-provisioning evidence, and an RPC-function claim would require a broader architectural comparison. The packet adjudicates none of those alternatives. Bypass failure is unresolved: technical availability does not prove effective exit, and unmeasured inconvenience does not prove its failure.

5.4 Discipline Burden

Restrictions and connectivity failures are plausible provocations, but the archive defines no complete event population and tracks neither switching, organized demands, nor responsive term changes. Provider switching may itself discipline an Infura-specific carrier even when the participant remains within hosted RPC infrastructure; discounting that response would make the carrier unfalsifiable. With no rule-bound provocation-and-response corpus, discipline failure is unresolved.

5.5 Verdict

Burden	Status	Reason
Admission	Unresolved	The documents show hosted access without local-node operation, but not a defined prior bottleneck and observed widening for a specified participant class.
Residual mediation	Supported as a component	The default wallet task traverses an RPC service outside consensus.
Dependence	Configured component only	The default structures the unreconfigured path; population reliance and Piersonian path dependence are unestablished.
Governance relevance	Supported as a component	Documented regional restrictions and connectivity conditions can shape default-path access and continuity, but cannot cure the Admission gap.
Failed bypass	Unresolved	Alternatives are documented, while participant-class feasibility has not been measured.
Failed discipline	Unresolved	The dossier lacks a complete provocation-and-response corpus.
Distributive capture	Unestablished	Admission and both conjunctive terminal burdens remain open.

Under the ordered protocol, the case stops at unresolved Admission. Independently, later component evidence identifies residual mediation, configured dependence, and governance relevance. Earlier drafts crossed the threshold using uncalibrated time, cost, and degradation numbers and two selected provocations; those moves are withdrawn.

Distributive capture therefore remains a draft diagnostic protocol. Its evidentiary stop is clear: mediation, concentration, defaults, and gatekeeping acts can make infrastructure governance-relevant; capture begins only after a fixed carrier defeats both practical exit and effective voice. The default Ethereum access path has not crossed that line on the evidence assembled here.

6. Dated Instruments and Their Limits

Concentration measurement remains useful for locating research sites. It does not adjudicate the default RPC path, and it cannot establish capture on its own.

Relayscan’s documented 24-hour MEV-Boost JSON endpoint captured on May 17, 2026 placed Titan at 49.17%, BuilderNet at 24.66%, and Quasar at 15.07% of builder extra-data share (Relayscan 2026). The source interval ran from May 16, 2026 at 03:26:37 UTC to May 17, 2026 at 03:26:37 UTC. The repository packet records a raw-response SHA-256 of 798b30e25ab02233d8e8851a2bbb0697ec636f357a535ed99f8d07acd7ff33bc; after deterministic redaction of transient response tokens, its manifest is pinned as 274b06d9400f029738382306001005d512edc81606801ab3ff7f4d3d3088afeb.

Those numbers prove a reproducible reading of one instrument over one window, subject to the source’s own classification and coverage. They do not prove durable market share, common ownership, coordination, censorship capacity, participant dependence, failed entry, failed bypass, or failed discipline. Builder concentration remains a candidate research site rather than an adjudicated application.

Ethernodes and Rated Network remain relevant source routes for node-geography and staking questions (Ethernodes 2026; Rated Network 2026). The May 17 direct Ethernodes routes returned Cloudflare challenge pages. Rated’s unauthenticated API returned an authorization boundary, and its explorer route returned a checkpoint. The evidentiary conclusion follows: the current local record does not support treating their May 2026 dashboard values as submission-grade numerics. No such values appear here.

The artifacts answer three different questions. Custody asks, What exactly did the researcher inspect? Measurement validity asks, What population and construct does the instrument represent? Causal analysis asks, Why does that reading bear on authority? A cryptographic hash strengthens only the custody answer.

7. Comparative Pressure and Withdrawn Extensions

Historical analogies clarify why distribution and control can grow together, and they enter here as genealogical prompts and nothing more.

Eisenstein’s (1979) account of print emphasizes the transformative consequences of reproducible texts, while the institutions of production, licensing, and circulation continued to organize access. Chandler (1977) and Hughes (1983) show how rail and utility scale depended on managerial coordination and system-building. Lessig (1999) identifies code and architecture as modalities of regulation in networked environments. Each literature makes a different causal argument. Their conjunction motivates the residual-infrastructure question; it supplies no cross-period threshold test.

The earlier draft treated Bitcoin’s full-node and signer paths as an in-domain negative case. That classification is withdrawn. Technical availability of local verification and custody cannot establish effective bypass for a matched participant class. The dossier has no comparative study using the same tasks, competence assumptions, service dimensions, and event rules applied to Ethereum. Bitcoin can become a future comparator after that design exists.

The proposed Solana/Firedancer prediction is also withheld. A testable prediction requires a dated, stake-weighted client-distribution baseline; a disclosed rule for classifying independent codebases; an adoption threshold justified before observation; a follow-up horizon; and predefined adjacent layers where relocation would be measured. The dossier reserved such a baseline without capturing it. Current deployment status cannot repair an ex ante prediction retroactively. No claim about Firedancer adoption or Solana’s deployed client distribution is made here.

These withdrawals keep the diagnostic vocabulary from borrowing authority from cases it has not adjudicated.

8. Study Required for Adjudication

The default access path is testable. The design should be registered before a new event supplies the outcome.

8.1 Participant Cohorts and Tasks

The first study should separate at least three cohorts: ordinary wallet users without node-administration experience, technically experienced self-custody users, and small application teams operating production endpoints. Each cohort should receive a task that preserves the same substantive capacity across paths, such as reading state and submitting a valid transaction under a declared security model.

The alternatives should be fixed at registration: the default Infura path, at least two independently operated hosted providers, a locally operated execution-and-consensus pair, and another wallet or client where task equivalence can be maintained. The carrier-level hypothesis should also be fixed: Infura, the hosted-provider market, or the RPC function.

8.2 Measures

The bypass study should record setup time, successful completion, recurring labor, direct and opportunity cost, latency, availability, error recovery, privacy exposure, security assumptions, and feature loss. Smart Transaction availability is one documented feature difference. The study should avoid treating every feature as equally necessary; task requirements and participant preferences must be elicited in advance.

Service floors should come from observed task requirements and baseline distributions. The analysis plan should specify equivalence or non-inferiority margins, sample size, uncertainty intervals, exclusion rules, and how failed setups are counted. A threshold derived after seeing which path wins would reproduce the arbitrariness this revision removes.

8.3 Event and Discipline Record

The discipline study should define a prospective event registry. Each qualifying restriction, outage, routing-policy change, material price change, or default-configuration change would receive a timestamped record. The registry should track provider switching, wallet reconfiguration, complaints through declared venues, competitor entry, policy revision, default revision, and persistence.

A discipline claim also needs a counterfactual. Provider reversal following organized pressure may indicate voice. Rapid switching may indicate exit. Entry by a rival may indicate market discipline. A term that persists amid low switching and organized, unsuccessful contestation is stronger evidence of failed discipline. Causal attribution will still require process evidence, since timing alone cannot distinguish response from coincidence.

8.4 Decision Rule

The protocol should classify each carrier at one of the five states in Section 4. A capture verdict would require:

a defined capacity, prior bottleneck, participant class, and observed widening sufficient for Admission;
governance-relevant conduct linked to the specified carrier;
failure of every registered substitutable path to meet the participant-class service floor;
at least one qualifying provocation with a complete response record;
failure of registered discipline mechanisms within the justified observation horizon;
robustness to plausible alternative explanations, including ordinary service differentiation and transaction-cost efficiency.

An Infura-specific capture claim fails if another hosted provider performs the task within the registered bounds. Provider switching that works while self-provisioning fails leaves a hosted-market claim open and shifts the burden to competition evidence. Organized voice that changes the default or the provider term defeats the discipline arm. These outcomes give the protocol a real possibility of rejecting its strongest claim.

9. Scope of Inference

Each finding reported here is bounded to its own participant path. The adjudicated path is narrow: a MetaMask user performing an Ethereum task through the wallet’s configured default provider. The period runs to official product materials as they stood on July 15, 2026, together with one earlier instrument reading covering May 16, 2026 at 03:26:37 UTC to May 17, 2026 at 03:26:37 UTC. Carrier level stays fixed at the named default provider; a hosted-RPC-market claim and an RPC-function claim describe different objects, admit different alternatives, and carry their own evidence burdens.

Within that boundary the article reports what official product documentation states about a default provider, specified regional restrictions, documented connectivity conditions, and configurable alternatives. Those materials support residual mediation, configured dependence, and governance relevance as component findings for the specified path, while Admission, effective bypass, and effective discipline stay unresolved and capture stays unestablished. One reproducible reading of a single builder-share instrument over one 24-hour window is reported outside that boundary; it locates a research site and does not adjudicate the default RPC path.

Outside that boundary sit population prevalence on the default path, measured substitution cost across competing paths, a complete provocation-and-response corpus, durable market share, common ownership, coordination, censorship capacity, deployed client distribution on any network, and the commercial motive or legal authorship behind any restriction. No inference here depends on them. Because the study collects no original on-chain data and inspects no code checkout, it also reaches no conclusion about deployed software, validator adoption, wallet configuration, or network behavior.

Several limits are already binding. The confirmed extension is empty. Path-specific analysis of MetaMask and Infura does not establish Admission. It draws on official documentation rather than usage telemetry. The July 15 packet is an acquisition observation log with reported fingerprints and selected extracts, not independently verifiable raw-page custody or a publication-date record. Relayscan’s result remains a stale 24-hour instrument reading, preserved for reproducibility rather than current market description. Ethernodes and Rated numerics remain excluded, historical cases are not process-traced under the protocol, and no Bitcoin comparator, Solana baseline, code checkout, or deployment observation supports the present claims.

10. Falsification

The affirmative claim here is ordinal: distributive capture names a five-state conjunctive threshold whose confirmed extension is empty, and the default MetaMask and Infura path supplies residual mediation, configured dependence, and governance relevance as later-state components while the ordered case stops at unresolved Admission. Four observations bear on that claim. One would defeat the ordered stop, one would defeat a component finding the claim rests on, and two would resolve burdens the article leaves open without disturbing the ordered stop, because a later component cannot cure an earlier gap. None of the four can be settled from the material assembled here, and each names what a reader would have to obtain.

A dated record of a prior bottleneck in a defined capacity, together with observed widening for a named participant class, would satisfy the Admission burden and defeat the stop reported in 5.5 Verdict and restated in 1. Claim and Evidentiary Stop. The documents examined here show hosted access without local-node operation and contain no such comparison, so this condition names an artifact the record does not currently hold.
Wallet telemetry recording who remains on the default path, for which tasks, and for how long would settle the Dependence question that 5.1 Residual Mediation and Dependence leaves at configured dependence, and telemetry showing that the specified class ordinarily moves off the default would defeat reliance for that class. The archive holds no wallet telemetry.
A preregistered comparison of the default provider path against at least two independently operated hosted providers, a locally operated execution-and-consensus pair, and another client, scored on the measures set out in 8.2 Measures, would resolve the bypass burden in either direction: a rival hosted provider completing the task within registered bounds would defeat the Infura-specific carrier hypothesis, while failure of every registered path at the participant-class service floor would score failed bypass. No participant experiment of that design appears in the record, so the finding in 5.3 Bypass Burden stands unresolved.
A provocation-and-response corpus built on a declared inclusion rule, covering every documented service restriction affecting the participant class over a fixed period along with the switching, contestation, provider response, default changes, entry by rivals, and term persistence that followed, together with the process evidence required for causal attribution in 8.3 Event and Discipline Record, would make the discipline arm scoreable and resolve the status recorded in 5.4 Discipline Burden. The dossier defines no complete event population, so no such corpus can be assembled from what is on hand.

Each condition names an obtainable artifact: a dated bottleneck record, wallet telemetry, a registered comparator study, or an event registry. None of them sits in the current record, which is why the ordered classification stops where it does and why later-state evidence is reported as components.

Three forms of falsification bear on the concept itself.

Empirical falsification occurs when governance-relevant residual paths remain practically bypassable or responsive to discipline under preregistered tests. A negative result is informative: default power and concentration can exist below capture.

Discriminant falsification occurs when the five-state protocol adds no classification power beyond obligatory passage points, gatekeeping, path dependence, or transaction-cost analysis. If every result can be stated with greater precision in those vocabularies, distributive capture is redundant.

Design falsification occurs when participant classes, carriers, or thresholds shift after outcomes are known. A concept that survives by moving its unit of analysis cannot be tested.

References

Barzilai-Nahon, Karine. 2008. "Toward a Theory of Network Gatekeeping: A Framework for Exploring Information Control." Journal of the American Society for Information Science and Technology 59(9): 1493-1512. DOI.

Callon, Michel. 1986. "Some Elements of a Sociology of Translation: Domestication of the Scallops and the Fishermen of St Brieuc Bay." In John Law, ed., Power, Action and Belief: A New Sociology of Knowledge?, 196-233. Routledge & Kegan Paul.

Carpenter, Daniel, and David A. Moss, eds. 2014. Preventing Regulatory Capture: Special Interest Influence and How to Limit It. Cambridge University Press. DOI.

Chandler, Alfred D. 1977. The Visible Hand: The Managerial Revolution in American Business. Harvard University Press.

De Filippi, Primavera. 2020. "Blockchain Technology and Decentralized Governance: The Pitfalls of a Trustless Dream." SSRN.

Edwards, Paul N. 2003. "Infrastructure and Modernity: Force, Time, and Social Organization in the History of Sociotechnical Systems." In Thomas J. Misa, Philip Brey, and Andrew Feenberg, eds., Modernity and Technology, 185-225. MIT Press.

Eisenstein, Elizabeth L. 1979. The Printing Press as an Agent of Change. Cambridge University Press.

Ethernodes. 2026. "Ethereum Mainnet Statistics: Countries." Living dashboard, captured May 17, 2026. Ethernodes.

Ferreira, Daniel, Jin Li, and Radoslava Nikolowa. 2023. "Corporate Capture of Blockchain Governance." Review of Financial Studies 36(4): 1364-1407. DOI.

Hirschman, Albert O. 1970. Exit, Voice, and Loyalty: Responses to Decline in Firms, Organizations, and States. Harvard University Press.

Hughes, Thomas P. 1983. Networks of Power: Electrification in Western Society, 1880-1930. Johns Hopkins University Press.

Infura. n.d. "Frequently Asked Questions." Living documentation, accessed July 15, 2026. Infura.

Jasanoff, Sheila, ed. 2004. States of Knowledge: The Co-Production of Science and the Social Order. Routledge.

Lessig, Lawrence. 1999. Code and Other Laws of Cyberspace. Basic Books.

Mann, Michael. 1984. "The Autonomous Power of the State: Its Origins, Mechanisms and Results." European Journal of Sociology 25(2): 185-213. DOI.

MetaMask Support. n.d.-a. "Why Infura Doesn’t Serve Certain Regions." Living documentation, accessed July 15, 2026. MetaMask Help Center.

MetaMask Support. n.d.-b. "How to Troubleshoot Infura Connectivity Issues in MetaMask." Living documentation, accessed July 15, 2026. MetaMask Help Center.

MetaMask Support. n.d.-c. "How to Add a Network." Living documentation, accessed July 15, 2026. MetaMask Help Center.

MetaMask Support. n.d.-d. "How to Use a Local Node with MetaMask." Living documentation, accessed July 15, 2026. MetaMask Help Center.

Musiani, Francesca. 2015. "Giants, Dwarfs and Decentralized Alternatives to Internet-based Services: An Issue of Internet Governance." Westminster Papers in Communication and Culture 10(1): 81-94. DOI.

Olson, Mancur. 1965. The Logic of Collective Action: Public Goods and the Theory of Groups. Harvard University Press.

Peltzman, Sam. 1976. "Toward a More General Theory of Regulation." Journal of Law and Economics 19(2): 211-240. DOI.

Pierson, Paul. 2000. "Increasing Returns, Path Dependence, and the Study of Politics." American Political Science Review 94(2): 251-267. DOI.

Rated Network. 2026. "Ethereum Staking Explorer." Living dashboard, captured May 17, 2026. Rated Network.

Relayscan. 2026. "MEV-Boost Builder Statistics." Living JSON endpoint, captured May 17, 2026. Relayscan.

Star, Susan Leigh, and Karen Ruhleder. 1996. "Steps Toward an Ecology of Infrastructure: Design and Access for Large Information Spaces." Information Systems Research 7(1): 111-134. DOI.

Stigler, George J. 1971. "The Theory of Economic Regulation." Bell Journal of Economics and Management Science 2(1): 3-21. DOI.

Williamson, Oliver E. 1985. The Economic Institutions of Capitalism. Free Press.