When the backup isn’t a backup

On 21 September 2026, a telecommunications failure at a Philadelphia air-traffic-control facility disrupted flights across the US Northeast. Reuters reported hours of restrictions at major airports before operations resumed later that day. [A1]

According to NBC New York’s account of FAA Administrator Bryan Bedford’s remarks, an older Verizon primary circuit failed. When the system attempted to switch to its backup, a fibre line was found cut by construction work in New Jersey. The report identified the backup cable as Verizon’s and said the primary circuit was due for replacement. [A7]

This describes the reported discovery sequence, not a verified chronology of both faults. The disruption does not by itself establish that aviation safety was compromised.

Philadelphia Terminal Radar Approach Control (TRACON) Area C has handled Newark traffic since its transfer from New York in 2024. The FAA’s log records telecommunications and equipment problems on 28 April 2025, an approximately 90-second communications and radar-display outage on 9 May, and a roughly two-second loss of radio frequencies on 19 May. [A4]

In May, the agency announced additional connections, fibre upgrades, temporary backup and a local STARS radar-data hub to reduce dependence on New York. Its log also records a separate 11 May incident during which traffic was slowed while redundancies were checked. [A4]

On 3 July 2025, the FAA announced a new fibre network between New York and Philadelphia on two separate communications paths. It explicitly reported testing during the preceding month. The statement also said a temporary satellite backup had been deployed and would remain for the time being. [A5]

A 25 September 2025 update still described the local STARS hub as being established, alongside the completed fibre transition and deployed satellite backup. This was a record of progress across different workstreams, not a declaration that every change was complete. [A6]

The sources reviewed here do not map the circuits involved in September 2026 to the network announced in July 2025. They do not establish whether the temporary satellite system remained available or could support the affected functions. [A5], [A6], [A7]

The precedent changes the question, not the verdict.

The same facility does not necessarily mean the same fault. A later outage can involve different links or functions, dependencies outside an earlier upgrade, or damage after successful testing. None of these possibilities is established here.

Nor is verification absent merely because a complete test record is not public. The relevant distinction is between work announced, work reported as delivered, testing reported, and evidence that a particular service can withstand a particular disruption.

That is the ABT question: can those responsible for the system trace the correction from the original problem through the delivered change, the test result and the remaining exposure?

Using the same carrier is a dependency to examine, not proof of a common cause. One provider can supply physically separate routes; different providers can share infrastructure. The assessment must examine the actual dependencies rather than count supplier names.

A diagram describes redundancy. Monitoring and realistic testing provide evidence of its condition. An incident tests that evidence under actual operating conditions.

Time

How long can the essential service remain safely available in degraded mode? How quickly can teams detect a loss of redundancy, activate an alternative and restore the necessary capacity?

The relevant clock starts when protection is lost, not merely when the passenger notices the disruption.

Trust

Confidence needs a defined scope: which service, configuration, load and failure scenario were tested; what passed; and what exceptions remained.

The FAA reported testing in 2025. The question is not whether a test existed, but how its scope relates to the affected service and what changed afterwards. [A5]

Cooperation

Who connects construction and maintenance information, telecommunications dependencies and operational decisions across organisational boundaries? Who can act when that information changes?

Useful assurance need not disclose sensitive network details publicly. It should, however, be available to authorised operators and oversight bodies.

These questions are the author’s ABT interpretation, not investigation findings attributed to the FAA, Reuters or NBC New York.

For a board or operating committee, I would ask six questions.

1. Which essential service must survive, for how long, and at what minimum capacity?
2. What plausible combination of failures could defeat its primary and backup arrangements, including shared dependencies?
3. When was failover last tested end-to-end under representative conditions, and what limitations remained?
4. How do we monitor backup availability independently of primary use, and who can restrict operations or stop work when protection is lost?
5. Following maintenance, construction or configuration changes, who confirms that the earlier test evidence still applies?
6. When a correction has an agreed deadline, who reports the delivered scope, verification results and unresolved risks, and who accepts those risks?

A restoration closes an outage.

Verification closes the correction loop.

That closure has a scope. Material changes, new damage or different operating conditions can require the evidence to be renewed.

This is commentary on the cited public sources, not an independent technical investigation.

Those sources do not establish that the 2025 programme failed, that the 2026 incident affected the upgraded paths, or that the satellite backup was absent or defective. They do not provide a complete component-level timeline or determine whether shared ownership contributed to the outage.

A public announcement is not a full assurance record. An incomplete public record is not proof of incomplete internal assurance.

The lesson is therefore a question to investigate, not a verdict on the adequacy of the FAA’s engineering or testing.

1. [A1]

   Reuters, David Shepardson. “Thousands of US flights snarled before FAA resolves telecom problem.” 21 September 2026.

   View source ↗

   Relevance: Reporting on the September 2026 disruption and resumption of operations.

2. [A4]

   Federal Aviation Administration. “FAA Statements on Newark Liberty International Airport.” 2025 newsroom log; relevant entries dated 28 April and 7, 9, 11, 13 and 19 May 2025.

   View source ↗

   Relevance: Newark airspace transfer; dated incidents; corrective workstreams; checks of redundancies during the 11 May incident.

3. [A5]

   Federal Aviation Administration. “Secretary Duffy Announces Completion of Another Major Milestone for Operations at Newark Liberty Airport.” 3 July 2025.

   View source ↗

   Relevance: Two-path fibre network, reported testing and temporary satellite backup.

4. [A6]

   Federal Aviation Administration. “FAA Extends Order Limiting Operations at Newark Liberty International Airport.” 25 September 2025.

   View source ↗

   Relevance: Subsequent dated description of completed and continuing workstreams. This is not evidence of their status in September 2026.

5. [A7]

   NBC New York, Jennifer Millman and Tom Shea. “NJ Transit blamed for air travel nightmare that halted flights across northeast for hours.” Published 21 September 2026. Accessed 22 September 2026.

   View source ↗

   Relevance: Reported account of the older Verizon primary circuit, severed backup cable and discovery at attempted switchover.

The interpretation, TTC framing and proposed board questions are Abdeldjellil Bouzidi’s commentary drawing on The Airport Bathroom Test: Why Strong Systems Break and What Builders Do Differently (2026).