*

*
Ricardo F. Morín
November 17, 2026
Oakland Park, Florida
*
Between November 12 and November 17, 2026, failures interrupted services that people ordinarily encounter as dependable parts of daily life. Microsoft users lost access to business communications. A malfunction in the Federal Aviation Administration’s telecommunications network led to a nationwide suspension of aircraft departures. United Airlines delayed flights when crews could not obtain required departure information. JPMorgan Chase encountered problems processing transactions, with delays reaching customers of other banks. Apple users then experienced interruptions involving account access and several of the company’s online services. These events occurred within days of one another, close enough to suggest a common disturbance. Their proximity, however, is the beginning of an inquiry, not evidence of a single cause.12345
The distinction matters because technological dependence does not operate at only one level. Two failures may have independent origins while producing similar public consequences. A failure originating within one company may affect services supplied by several others. Multiple services may cease working because they depend upon the same internal component, although their users experience them as separate products. Alternatively, organizations with no shared faulty component may be susceptible to the same kind of weakness because they have adopted comparable practices of consolidation, automation, and recovery. To call all these conditions interconnected is not incorrect, but it is insufficiently discriminating to explain what failed.
The Microsoft disruption provides an unusually revealing instance of dependence within a single technological environment. Its preliminary account identified a latent software defect introduced months earlier, additional traffic following a Teams-related failover, and reduced available server capacity. In combination, these conditions overwhelmed networking gateways used by several Microsoft 365 services. Bringing servers back did not immediately restore service because the waiting requests could overwhelm the recovering capacity again. Measures intended to maintain availability thereby became implicated in the spread or persistence of the interruption. Microsoft’s explanation identifies an interaction among conditions rather than a single defective component acting in isolation. It remains a preliminary account, but it supplies a documented pathway from local weakness to widespread unavailability.16
The FAA interruption arose in a different environment. The agency reported that unusual traffic associated with telecommunications equipment at El Dorado, Arkansas, and Amarillo, Texas, made its FAA Telecommunications Infrastructure network unstable. The disturbance reached 17 of its 22 en route air traffic control centers, although not every communications or surveillance function was lost. An investigation was still required to determine precisely how replacement equipment, configurations, and network behavior interacted. The national ground stop was a response to uncertainty about the reliability of services necessary for safe aircraft movements. The significant question is how difficulties associated with a limited number of facilities could affect so many operating centers, and whether safeguards expected to contain such disturbances had been adequately tested.2
Microsoft and the FAA did not thereby suffer the same failure. The documented triggering circumstances differ, as do the technologies, responsibilities, and operational decisions involved. Yet both cases draw attention to an important possibility: a mechanism designed to distribute work or maintain continuity can transmit stress when the conditions that support it cease to hold. In Microsoft’s case, traffic redistribution and limited capacity are part of the published account. In the FAA case, the technical inquiry still has to establish the pathway through which abnormal traffic reached so much of the network. The similarity is presently diagnostic, not proof of a common engineering defect.
The interruption at United Airlines was different again. Its pilots’ union reported that the airline could not supply departing aircraft with final weight information, and flights were delayed while the company addressed its technology problem. Aircraft already in flight were not reported to have been affected. Here the immediate concern was not that all aircraft functions had failed, but that a condition necessary for safe departure could not reliably be verified through the ordinary process. Technology had become indispensable to the performance of a particular task. Whether an alternative procedure could satisfy that requirement depended on applicable operational and safety standards.3
The JPMorgan Chase disruption illustrates a more direct connection among organizations. The bank attributed delayed transaction processing to a software problem in its own systems. JPMorgan Chase acknowledged that delays could reach other banks when they were parties to transactions involving Chase, while customers also reported difficulties involving other financial institutions and payment services. A delay in a transaction moving between institutions does not establish that both institutions experienced an internal malfunction. An originating failure can become visible at every point where another institution depends upon the delayed service. The distinction between an affected customer and a failed provider is therefore essential: the consequences may be shared even when the technical fault is not.4
Apple’s interruption offers a related distinction within a company’s own range of services. Reports on November 17 described difficulty signing into Apple Accounts and using iCloud, the App Store, Apple Music, and other functions. Apple subsequently attributed the disruption to an internal update and said its services were being restored. An identity or account-access function can affect multiple activities experienced by customers as distinct, though the precise internal propagation of this incident has not been publicly reconstructed in the reports examined. Apple’s stated cause provides no evidentiary bridge to the FAA, Microsoft, United, or Chase. What its interruption adds is another example of the difficulty of perceiving a service’s hidden dependencies from the apparent simplicity of its public interface.5
The succession of failures consequently requires more than a choice between coincidence and coordination. One category consists of independently initiated incidents whose dates happen to be close. Another consists of failures propagated through an actual dependency, as when transaction delays originating in one bank affect customers dealing with another. A third consists of distinct services impaired by a shared component or a common operating environment, as Microsoft’s account describes. A fourth concerns recurring vulnerabilities across otherwise separate systems: inadequate testing under realistic loads, insufficiently independent backup arrangements, concentrated service provision, or recovery procedures that fail under conditions different from those for which they were designed. The fourth category can be consequential without supplying a shared cause for the particular incidents under examination.
The temptation to seek one hidden origin is understandable. A traveler unable to depart, an employee unable to receive messages, a customer unable to complete a payment, and an individual unable to authenticate an account encounter the same immediate fact: something essential is beyond their control. Because the failures resemble one another at the point of use, their causes may also appear to belong together. But the user’s experience is evidence of interruption, not a technical map of its origin. Similar consequences are consistent with very different causal histories.
Nor does a concentration of news reports establish that the underlying rate of failure has suddenly increased. To make that judgment would require a baseline: the frequency, duration, severity, and reporting of comparable interruptions before these six days, adjusted for changes in the number of users and services. A cluster may reflect an unusual concentration of events, heightened attention after the first emergency, or both. The relevant question is not whether the reports feel unprecedented but what comparison would demonstrate that they are. Public anxiety is a genuine consequence of recurrent disruption; it is not itself evidence of a common technical origin.
The more important question concerns the terms under which indispensable services have come to depend upon one another. Consolidation can make systems efficient and affordable by allowing many activities to use the same computing resources, networks, authentication systems, suppliers, or operating processes. But economies achieved through shared components also concentrate consequences when those components fail. An apparently separate service may be separate in its commercial presentation while remaining dependent on a facility, provider, or process that it cannot independently replace. The boundary familiar to a customer is therefore not necessarily the boundary that matters to an investigator.
The difficulty is not confined to shared ownership. Two organizations may use different networks but depend upon the same electricity supply, identity provider, payment intermediary, software library, or maintenance contractor. They may also use independent equipment configured according to identical assumptions about traffic and recovery. In one case a common resource can transmit a disturbance; in another a common design weakness can cause separate disturbances under comparable stress. These possibilities cannot be treated as actual connections without evidence showing that the relevant dependencies existed at the time. They identify what an investigation should look for, not what this particular week has established.
The Bank of England’s work on operational resilience provides a useful account of why such distinctions matter. It describes several ways in which a disturbance can move beyond the organization where it begins: relationships among firms, concentration in a small number of service providers, common technologies, and the difficulty of replacing a critical service quickly. These characteristics can amplify an otherwise local incident. They also complicate the attribution of responsibility, because the institution suffering an interruption may not control the component on which its service depends. The analysis is addressed to financial stability, but the underlying problem of indispensable external dependencies is not peculiar to banking.7
A claim that two failures share a cause should therefore be tested through an identifiable chain of evidence. Investigators would need to establish when each service first became impaired, which components failed before others, what changes or loads preceded the failures, and which connections were carrying information or work between the affected functions. They would need configuration histories, time-aligned operational logs, incident records, and evidence of what occurred when traffic was rerouted or backup systems were activated. A dependency diagram could disclose whether apparently separate services relied on the same upstream function. The evidence would then have to show not merely that the common function existed, but that its behavior could account for the interruptions observed.
A shared supplier alone would not prove a shared failure. A common software product would not prove that the same defect was activated. Near-simultaneous loss of service would not prove that one outage caused another. Conversely, different immediate faults would not exclude a common vulnerability if the same testing assumptions or recovery practices made those faults consequential. Each proposition requires a different inquiry. Failure to distinguish them can produce opposite errors: an unsupported allegation of coordination or a premature declaration that independent initiating events reveal nothing about the reliability of the systems in which they occurred.
The distinction between cause and propagation is particularly important for the public. The cause identifies the condition that initiated a disturbance. Propagation explains why that disturbance reached other functions or institutions. Recovery shows whether the system could continue performing an indispensable task, through alternative arrangements, while the disturbance was being contained. An account limited to the failed router, software bug, or internal update may answer the first question while leaving the others unresolved. A complete account need not prove that every participating institution caused the incident. It must show how their dependencies altered its consequences.
These records are seldom available to the people most affected. Customers see applications, departure boards, account balances, or status messages. Operators see internal systems; contractors may control portions of the relevant networks; and several organizations may each possess only a segment of the evidence. Even published incident reports sometimes describe the immediate fault more fully than the underlying dependencies that made it consequential. The public may thus know that services were restored without knowing whether the restoration addressed the conditions that permitted the disruption to spread.
This disparity creates a legitimate demand for disclosure, but not necessarily for unrestricted publication of sensitive network diagrams or security configurations. Detailed records may need to be available to independent technical reviewers and public authorities empowered to examine them, while public explanations identify the affected functions, the responsible parties, the classes of dependency involved, the safeguards that operated or failed, and the corrective measures whose effectiveness will be tested. Confidentiality can protect a system against exploitation. It cannot by itself establish that the system’s failures have been adequately understood or that future interruptions are less likely to have comparable consequences.
There is also a question about what counts as an adequate safeguard. A backup is not independent merely because it is called a backup. If primary and secondary paths depend on the same power supply, authentication service, configuration process, or traffic assumptions, their apparent redundancy may disappear under a common disturbance. A recovery procedure that works when one server fails may not work when many servers restart under accumulated demand. A replacement component that passes isolated testing may still behave unexpectedly in the full operating network. Reliability therefore has to be assessed under the conditions that cause ordinary safeguards to be needed, not solely under the conditions in which those safeguards appear satisfactory.
The National Institute of Standards and Technology describes cyber resiliency as the capacity to anticipate, withstand, recover from, and adapt to adverse conditions, stresses, attacks, or compromises involving cyber resources. The Cybersecurity and Infrastructure Security Agency likewise treats dependencies and cascading disruptions as matters for infrastructure planning, not merely as technical afterthoughts once a service has failed. These approaches shift attention from the apparent health of each component to the continued delivery of the function that the components jointly support. That distinction is decisive when a service is indispensable: what matters is not whether its equipment can be described as operational in isolation, but whether the public can still rely on the service under foreseeable disturbance.89
Indispensability also changes the assignment of responsibility. Individuals and businesses can sometimes maintain alternative communications, payment methods, or work procedures, but the ability to improvise cannot be made the principal guarantee of public continuity. Some activities admit temporary substitutes; others must be suspended until their safety requirements can be met. Providers and regulators must determine which alternatives are genuinely independent, which are available at the necessary scale, and who bears the costs when they fail. A system cannot be called resilient merely because its interruption was survived by those obliged to endure it.
The events of November 2026 do not presently demonstrate a coordinated attack, a shared technical origin, or a single actor responsible for all the disruptions. They do demonstrate several different ways in which complex services can become unavailable and several different degrees of connection between an originating fault and its public consequences. Microsoft’s gateways illustrate failures interacting inside a shared environment. Chase’s transactions show how consequences can cross institutional boundaries. The FAA raises questions about the reach of a disturbance through a national operating network. United demonstrates the suspension of an activity when required information is unavailable. Apple’s interruption makes visible a dependence on account and service functions normally taken for granted. Taken together, these are grounds for investigating the conditions of dependence, not for assigning the incidents a collective intention.
The diagnosis may never produce one comprehensive explanation, because there may be no single event to explain. It can nevertheless establish whether each institution understood the services on which it depended, whether safeguards were tested against plausible combinations of failure, and whether the public received an account proportionate to the consequences it experienced. Independence of immediate causes does not make shared vulnerability irrelevant; shared vulnerability does not establish a coordinated cause. The public needs the evidence to distinguish those conditions precisely because indispensable services are experienced as separate assurances while their continued operation may depend upon arrangements that remain unseen. The question is no longer only what failed. It is what had to remain functioning for everything else to work, who knew that dependence, and what was done to make it bearable when one part ceased to function.
Endnotes
- The Guardian, November 16, 2026: Microsoft identifies a latent defect, traffic redistribution, and capacity losses. The preliminary account describes interacting conditions within Microsoft 365; it does not connect the outage to other organizations’ incidents. Return to text
- FAA, November 15, 2026: Nationwide Ground Stop, Questions and Answers. The FAA reports unusual network traffic and 17 affected en route centers, while identifying its technical investigation as preliminary. Return to text
- CBS News / Associated Press, November 16, 2026: United Airlines technology outage. The pilots’ union identified missing final weights for departing aircraft. United reported no effects on flights already in progress. Return to text
- Reuters, November 16, 2026: JPMorgan Chase transaction delays. Chase attributed the problem to a software bug and acknowledged possible transaction delays affecting other banks, without identifying independent failures in their systems. Return to text
- Reuters, November 17, 2026: Apple attributes disruptions to an internal update; San Francisco Chronicle, November 17, 2026: Apple users report multiple interrupted services. The initial public accounts establish affected services and Apple’s stated cause; they do not reconstruct every internal technical dependency. Return to text
- University of Pennsylvania, Microsoft 365 service-health messages, November 12–13, 2026. The contemporaneous Microsoft notices describe server health-check failures, capacity problems, traffic redistribution, and difficulties restoring service. Return to text
- Bank of England, 2024: Operational resilience in a macroprudential framework. Distinguishes interconnections, common vulnerabilities, concentration, and difficulty substituting critical providers. Return to text
- NIST, SP 800-160 Volume 2, Revision 1: Developing Cyber-Resilient Systems. Establishes engineering principles for anticipating, withstanding, recovering from, and adapting to adverse conditions, stresses, attacks, or compromises involving cyber resources. Return to text
- CISA, Critical Infrastructure Security and Resilience. Addresses interdependent infrastructure and cascading consequences that can cross sector boundaries. Return to text