๐ Strawberry Tech Notes
Imagine thousands of aircraft preparing to depart across the United States.
Airports are operating. Pilots are ready. Passengers are boarding. Airlines have their schedules, aircraft, and crews organised.
Then, a problem develops in the telecommunications network used by air traffic control.
Departures across the country are paused.
This isn't a fictional disaster scenario. On November 13, 2026, the Federal Aviation Administration (FAA) issued a nationwide ground stop after a technical failure disrupted parts of its air traffic control communications network.
The disruption lasted nearly five hours. Thousands of flights were affected, and the consequences continued long after the original technical issue was addressed.
Just three days later, United Airlines experienced a separate computer outage that delayed hundreds of flights.
These incidents involved different organisations and systems, but they raise some important engineering questions.
- How can a problem at one remote network location affect operations across an entire country?
- Why can replacing old infrastructure introduce new risks?
- What happens when a critical system becomes unstable?
- And how do you recover when the software is working again, but the operation around it is still disrupted?
Let's break it down from a systems-reliability perspective.
1. What Actually Happened?
On Friday, November 13, the FAA experienced technical problems affecting communications and radar feeds used by air traffic control.
At approximately 3:30 p.m. Eastern Time, the FAA issued a nationwide ground stop.
A ground stop means aircraft are temporarily prevented from departing for specified destinations or, in this case, across the national system. It does not mean every aircraft already in the air was suddenly grounded.
Aircraft already airborne continued to land safely. According to the FAA, approximately 3,500 aircraft landed during the disruption.
The ground stop was lifted at approximately 8:15 p.m. Eastern Time.
That's nearly five hours.
But the technical problem was only the beginning of the disruption.
Flights had been delayed or cancelled. Aircraft were no longer necessarily where airlines expected them to be. Crews were displaced from their planned schedules, and passengers were stranded at airports.
The FAA reported approximately 12,462 flight cancellations on Friday. ABC News subsequently reported more than 14,000 cancellations across Friday and Saturday.
The incident demonstrated how a failure in one part of a national infrastructure system can affect thousands of operations that depend on it.
2. The Technical Failure: How Did One Router Cause So Much Trouble?
This is the most interesting part from an engineering perspective.
According to the FAA and the US Department of Transportation, the incident occurred during routine maintenance at a remote FAA communications location.
The work involved transitioning telecommunications infrastructure from older copper-based connections to fibre.
The contractor involved was L3Harris, the FAA's network service provider.
The transition involved moving from older Time-Division Multiplexing (TDM) technology to Internet Protocol (IP)-based communications.
Let's understand what that means.
TDM vs IP
TDM (Time-Division Multiplexing) is a technology that allows multiple communication channels to share a transmission line by assigning each channel a specific time slot.
Think of it as a road where each vehicle gets an allocated time to use the lane.
IP (Internet Protocol) is the foundation of modern packet-based networking.
Instead of reserving a fixed time slot for each communication stream, information is divided into packets that travel across networks using addressing and routing.
Modern IP networks provide flexibility, scalability, and efficient ways to connect systems.
The transition from older telecommunications infrastructure to modern IP networks is common across many industries.
However, changing the underlying network technology of a critical system requires careful planning and testing.
What went wrong?
The US Transportation Secretary explained that, during the transition, a router-related problem caused excessive connection attempts to flood other routers in the FAA's network.
The FAA later reported that its network was receiving approximately 1,000 connection attempts every 4โ5 seconds.
That's an enormous number of connection attempts being generated repeatedly.
The excessive traffic caused instability in the legacy FAA telecommunications network, affecting aircraft communications and radar feeds.
The FAA identified two potential locations associated with the excessive data.
- A radar site in Amarillo, Texas.
- A radio communications site in El Dorado, Arkansas.
The initial investigation connected the issue to work involving a new router at the El Dorado site.
The FAA and L3Harris isolated the affected locations, and the network stabilised.
The FAA also disabled the broadcasting functionality on the new routers as a precaution and applied the same precaution to approximately 60 similar routers.
An important distinction: The exact underlying fault had not been conclusively established in the FAA's November 15 update. Officials said they were still investigating whether the issue involved the router itself, installation work, or another equipment or configuration problem.
So we shouldn't confidently label this a specific software bug or a confirmed broadcast storm.
What we do know is that excessive network traffic destabilised legacy infrastructure and affected systems that air traffic controllers depend on.
3. Why Did the FAA Stop Departures Nationwide?
Here's where system reliability becomes particularly important.
Air traffic control depends on several systems working together.
Controllers need reliable information about aircraft positions, communication channels, flight movements, and other operational details.
Imagine that a system responsible for moving information between different locations becomes unstable.
Even if some radar feeds and communication frequencies remain operational, the overall picture may become less reliable.
The FAA reported instability at 17 of its 22 Air Route Traffic Control Centers.
These centres are responsible for managing aircraft travelling through large sections of US airspace.
If information cannot be transmitted reliably, controllers may not have the consistent information they need to safely coordinate normal operations.
This creates a serious problem.
Should the system continue operating with incomplete or unreliable information?
Or should operations be restricted until the system is stable?
The FAA chose the second option.
It issued a nationwide ground stop out of an abundance of caution.
This is an example of a fail-safe approach.
When a critical system cannot provide the level of reliability required for safe operation, the system moves into a safer restricted state rather than continuing as if nothing happened.
And there's an important point here:
The ground stop wasn't evidence that every communication and radar system had completely failed. The FAA explicitly said some frequencies and radar feeds remained operational.
It was the instability and uncertainty across critical parts of the network that made normal operations unsafe to continue.
All aircraft landed safely.
That matters more than keeping a departure schedule intact.
4. The Real Problem: A Small Failure Can Have a Large Blast Radius
Let's look at this through a software architecture lens.
Imagine an application that depends on several services.
Main Application
|
---------------------
| | |
Service A Service B Service C
| | |
---------------------
|
Shared Network
If the shared network becomes unstable, multiple services can experience problems at the same time.
The services themselves might be functioning correctly.
But they cannot reliably communicate with one another.
This is an important distinction between component reliability and system reliability.
A component can be healthy while the larger system is unavailable because one of its dependencies has failed.
The FAA incident illustrates a similar architectural risk.
A problem at a remote location affected network infrastructure used by multiple air traffic control facilities.
The FAA reported instability across 17 of its 22 Air Route Traffic Control Centers.
This is a large blast radius.
In software engineering, blast radius describes how much of a system can be affected when one component fails.
A well-designed distributed system tries to limit that impact.
For example
- A failure in one service should not automatically bring down every other service.
- Network traffic from one component should not overwhelm unrelated components.
- Critical operations should have independent fallback paths where possible.
- A local configuration problem should not unnecessarily become a global outage.
The challenge is that critical infrastructure often has complex dependencies that have evolved over decades.
Modern applications are not immune to this problem either.
A single shared database, authentication provider, DNS service, network gateway, or message broker can become a major point of failure if the architecture depends on it too heavily.
The lesson isn't that every component must be duplicated.
It's that engineers need to understand what happens when a dependency becomes unavailable or unreliable.
5. The Hidden Risk of Modernising Legacy Infrastructure
One of the most interesting aspects of this incident is that the failure happened during an infrastructure upgrade.
The work was intended to replace ageing copper telecommunications lines with fibre.
That is a modernisation effort.
But modernisation doesn't automatically make a system more reliable.
In fact, the transition period can introduce new risks.
Consider what happens when an organisation has infrastructure that has been operating for decades.
Over time, it accumulates
- Legacy hardware
- Older communication protocols
- Existing network configurations
- Dependencies between different systems
- Operational procedures built around the old environment
- Integrations that may not be fully documented
Replacing one component doesn't happen in isolation.
A new router may interact with older routers, legacy network protocols, existing routing rules, and systems that were never designed for the new configuration.
This is why infrastructure modernisation requires more than purchasing newer hardware.
It requires understanding the entire environment.
The observability challenge
Modern network environments often provide detailed telemetry, traffic analysis, and centralised monitoring.
But legacy infrastructure doesn't always offer the same level of visibility or integrate cleanly with newer monitoring tools.
When old and new technologies operate together, engineers may need to correlate information from multiple systems to understand what's happening.
That makes troubleshooting more difficult, particularly when unexpected traffic patterns develop across network boundaries.
In the FAA incident, excessive connection attempts were identified as a major part of the problem.
It is a useful reminder that monitoring network behaviour and understanding how components interact are just as important as upgrading the hardware itself.
What could have helped?
From a systems-engineering perspective, several practices are particularly important
1. Test in an isolated environment
New equipment and configurations should be tested against realistic traffic and compatibility scenarios before being introduced into production.
2. Use staged rollouts
Instead of changing multiple parts of a critical network simultaneously, introduce changes in controlled stages.
3. Have a rollback strategy
If a deployment causes unexpected behaviour, engineers should be able to return to a known stable configuration quickly.
4. Monitor traffic and network behaviour
Connection rates, routing behaviour, latency, packet loss, and unusual traffic patterns should be monitored during and after changes.
5. Define clear change windows
Maintenance should consider operational demand and the potential impact of failure.
The FAA said major air traffic control maintenance is typically performed at night, when traffic is lower. This particular replacement was at a remote location and was part of work that had been performed routinely for years.
Following the incident, the FAA said it would reevaluate how such maintenance is scheduled.
The bigger lesson is that a routine change is not necessarily a low-risk change.
The impact depends on the system being changed and the dependencies connected to it.
6. Why Didn't the Recovery End When the Network Came Back?
This is another important concept: technical recovery and operational recovery are not the same thing.
The FAA stabilised its network and resumed air traffic at approximately 8:15 p.m.
But airlines couldn't instantly return to their original schedules.
Why?
Because an airline is a complex, interconnected operational system.
A flight isn't just an aircraft travelling from one airport to another.
It depends on several things being available at the right time,
- The aircraft
- The pilots
- Cabin crew
- Ground staff
- Airport gates
- Baggage handling
- Fuel and maintenance arrangements
- Passenger connections
- The next scheduled flight
Let's look at a simplified example.
An aircraft is scheduled to fly
New York โ Chicago โ Dallas โ Miami
Now imagine the first flight is delayed by several hours.
The aircraft reaches Chicago late.
Its next departure is delayed.
Passengers miss connecting flights.
Crew members may approach their legally permitted duty limits.
The aircraft may no longer arrive in Dallas at the time required for its next scheduled service.
One delay can affect several later flights.
This is called a cascading operational disruption.
Airlines must then work to
- Reposition aircraft
- Reassign crews
- Rebook passengers
- Adjust flight schedules
- Manage missed connections
- Coordinate airport resources
American Airlines, for example, said it was working to reposition aircraft and crews following the FAA ground stop.
This is why a technical outage lasting a few hours can produce disruptions lasting much longer.
The system may be technically available again, but the operational state has already changed.
In distributed systems, restoring a service doesn't automatically restore every dependent process to its previous state.
That requires reconciliation, rescheduling, and coordination.
7. A Second Incident: United Airlines' November 16 Outage
Just three days after the FAA ground stop, United Airlines experienced another technology disruption.
This was a separate incident. It wasn't a continuation of the FAA network failure.
On Monday, November 16, United experienced an outage affecting applications used in its flight operations.
Pilots reported problems accessing information related to final aircraft weight, along with other operational tools.
The issue lasted approximately one hour and 40 minutes.
FlightAware recorded around 600 delayed United flights that evening, although it was unclear how many were directly caused by the outage.
Why did a problem involving aircraft weight information delay departures?
Because aircraft weight and balance are essential to safe flight operations.
Why does aircraft weight matter?
Before an aircraft takes off, pilots need accurate information about its weight and balance.
This involves information such as
- Aircraft weight
- Passenger load
- Baggage
- Cargo
- Fuel
- Centre of gravity
These values are used in takeoff performance calculations.
They help determine whether the aircraft can safely take off under the current conditions, including runway length, weather, and other performance factors.
If the required information cannot be verified, the aircraft cannot simply proceed as normal.
The outage affected the availability of information used in this process.
United delayed departures until the issue was resolved.
Flights already airborne and United Express regional flights were not affected.
United also said there was no indication that the incident was related to a cybersecurity issue.
The airline did not immediately identify the precise system responsible.
The engineering lesson
This is a different type of failure from the FAA incident.
The FAA incident involved telecommunications instability across air traffic control infrastructure.
The United incident involved the availability of airline operational applications and data.
But both demonstrate something important
When software provides information required for a safety-critical decision, availability and correctness matter more than convenience.
A system returning an incorrect weight calculation would be far more dangerous than a system temporarily refusing to provide a calculation.
A responsible system should not silently substitute unreliable information just to keep operations moving.
In safety-critical environments, stopping and verifying can be the correct response.
8. What Should Engineers Learn From These Incidents?
These events are worth studying because the lessons apply far beyond aviation.
Whether you're building a banking platform, a healthcare application, a payment gateway, or a distributed cloud service, the same reliability principles matter.
1. Design for failure, not just success
Developers often spend most of their time implementing the expected workflow.
But production systems also need to account for
- Network failures
- Unavailable dependencies
- Invalid data
- Unexpected traffic
- Hardware problems
- Partial outages
- Configuration mistakes
Ask yourself
_What happens if this dependency becomes unavailable?
What happens if it responds slowly?
What happens if it sends unexpected data?
What happens if it starts receiving far more requests than expected?_
These questions should be part of architecture and testing, not just something considered after an outage.
2. Understand your system's dependencies
A service may look independent in a diagram, but it could rely on shared infrastructure underneath.
For example
Application A โโโ
|
Application B โโโผโโ Shared Database
|
Application C โโโ
If the shared database fails, all three applications may be affected.
The same principle applies to network infrastructure, identity providers, DNS, cloud services, and message queues.
Knowing your dependencies is essential to understanding your system's real failure boundaries.
3. Control the blast radius
Avoid allowing one faulty component to overwhelm the entire system.
Depending on the architecture, techniques may include
- Rate limiting
- Network segmentation
- Circuit breakers
- Bulkheads
- Traffic isolation
- Resource quotas
- Redundant communication paths
For example, a circuit breaker in a distributed application can stop repeated requests to a failing dependency, preventing unnecessary pressure on an already unhealthy service.
These techniques don't eliminate every failure, but they can reduce how far the impact spreads.
4. Treat deployments as potential failure events
Even a routine infrastructure change can introduce unexpected behaviour.
A deployment strategy should include
- Pre-deployment validation
- Staged rollouts
- Monitoring
- Rollback procedures
- Clear ownership
- Post-deployment verification
The objective isn't simply to deploy successfully.
It's to detect unexpected behaviour quickly and limit its impact.
5. Build meaningful observability
When a production system fails, engineers need to understand what is happening.
That requires more than checking whether a server is running.
Useful observability includes
- Metrics
- Logs
- Distributed traces
- Network traffic monitoring
- Dependency health
- Error rates
- Latency
- Resource utilisation
For the FAA incident, understanding unusual connection rates and network instability was important to isolating the affected locations.
In an application, the equivalent might be identifying a sudden increase in requests, a failing dependency, or an unexpected change in response times.
You cannot reliably troubleshoot what you cannot observe.
6. Have a safe fallback
A fallback should not mean blindly continuing with incomplete or unreliable data.
Sometimes the correct fallback is
- Switching to a verified backup
- Moving to a degraded but safe operating mode
- Queueing requests for later processing
- Temporarily disabling a non-critical feature
- Stopping an operation until reliable information is available
The right fallback depends on the system.
For a social media application, temporarily disabling a feature may be acceptable.
For an aviation system, an unsafe fallback is not an option.
Graceful Degradation vs Fail-Safe Behaviour
Not every system should respond to failure in the same way.
Consider a typical web application.
If its recommendation service becomes unavailable, the application might still allow users to browse products, view their accounts, and complete purchases. It simply disables recommendations temporarily.
This is called graceful degradation: maintaining essential functionality while reducing non-critical features.
But consider a safety-critical system.
If an aircraft's required weight and balance information cannot be verified, the system shouldn't simply use an estimated value to keep operations moving.
The correct response may be to stop the operation until reliable information becomes available.
This is fail-safe behaviour.
The difference is important
- Graceful degradation: Keep operating with reduced functionality when it is safe to do so.
- Fail-safe behaviour: Move into a safe state when continuing normal operation could create unacceptable risk.
These approaches aren't mutually exclusive. Safety-critical systems can also have carefully designed degraded operating modes. But those modes must be validated and safe, not improvised during an outage.
The goal isn't always to keep a system running.
Sometimes the goal is to know when it must stop.
9. The Bigger Picture: Reliability Is an Infrastructure Problem
We often talk about modern technology in terms of new features.
AI models are getting bigger. Applications are becoming more sophisticated. Cloud platforms are providing more computing power.
But underneath all of this, there is infrastructure that must remain dependable.
Networks, routers, switches, databases, storage systems, communication channels, and hardware still matter.
A modern application can have beautifully designed software and an excellent user interface.
But if a critical network dependency becomes unstable, the application may still become unavailable.
The same is true for national infrastructure.
The FAA incident also highlights the challenge of modernising systems that cannot simply be switched off while upgrades are performed.
Aviation, healthcare, banking, energy, and telecommunications all operate systems where reliability is not an optional feature.
Modernisation must happen while maintaining the safety and continuity of existing operations.
That's a difficult engineering problem.
And it deserves more attention than simply asking why an old router was still in use.
A system isn't reliable just because it works when everything goes right. It's reliable when it can handle failure, contain its impact, recover safely, and protect the people depending on it.
That's what good engineering is supposed to do.
Sources and further reading
- FAA โ Nationwide Ground Stop: Questions and Answers (November 15, 2026)
- US Department of Transportation โ Update on Restoration of Airspace (November 14, 2026)
- Reuters โ US flight shutdown exposes deepening air traffic network crisis (November 13, 2026)
- ABC News โ FAA ground stop live updates (November 15, 2026)
- Associated Press โ United Airlines system outage delays departing flights (November 16, 2026)
Top comments (1)
This one hits close to home โ I've done copper-to-fiber cutovers in live racks, and the scariest part is exactly what happened here: the new gear works fine in isolation, but the legacy side sees traffic it was never built to handle. A router retrying a thousand times a second looks healthy on its own dashboard while it's flooding everything downstream. The part I'd underline is the change window: a staged rollout with a tested rollback would have kept this a ten-minute incident instead of a five-hour ground stop.