AKAWAN NEWSLETTER
Before leaving, let's stay in touch.
Once a month, receive practical insights from our AI and digital experts - covering use cases, real-world experiences, and trends - with no jargon, just useful ideas.
Your address remains confidential. You can unsubscribe at any time with a single click.
Ten hours without network
Wednesday, October 21, 2026, 11:20 AM: Orange and Sosh's mobile network goes down across France. Calls, SMS, data: nothing at all, for nearly ten hours. For the general public, it's an annoying day. For a factory, it's a much more uncomfortable question: what in your operations hangs by a single thread that you don't control?
What happened: a cutover, not an attack
The first reports came in around 11:20 AM, and service was only restored in the evening, around 8:30 PM (Sortiraparis / Selectra). Around noon, Downdetector recorded more than 11,000 reports (Planète Grandes Écoles).
Orange acknowledged an equipment failure in its mobile core network—that is, the central part that authenticates devices and routes all traffic (France 3 Occitanie). The specialized press, led by Numerama, pointed to a failed migration of part of the subscriber base to the new 5G SA core. Orange has not officially confirmed this lead (Tech Insider).
A revealing detail: according to the same source, restoring service required the equipment vendor's support to implement a temporary fix. And just the day before, Orange had shut down its 2G network in the last 43 departments where it was still operating. No technical link has been established between the two events, but they tell the same story: critical infrastructure undergoing modernization, where every cutover is a moment of vulnerability.
Why an industrial director should care
On the shop floor, the mobile network isn't just for phone calls. It often supports, without anyone having formally decided it, functions that directly impact production:
On-call duties and alarms: a machine alert sent by SMS to the on-call technician never arrives. The drift is only discovered at the shift change, not when it begins.
Remote maintenance: the vendor providing remote support via a 4G router can no longer connect. The downtime lasts as long as it takes for a technician to travel on-site.
Connected devices and M2M sensors: SIM cards in PLCs, meters, or monitoring units stop transmitting their data. Your dashboards show green because they simply haven't received anything new.
Remote sites: a warehouse or satellite facility connected solely via 4G becomes blind, and sometimes deaf.
The problem isn't that an outage happens: no carrier can make it impossible. The problem is not knowing in advance what it cuts off in your operations.
October 21, seen from the shop floor
To make things concrete, imagine a precision engineering SME of 120 employees, with a main plant and a small storage site 30 kilometers away. The scenario is fictional. The outage timeline, however, is real.
11:20 AM: the mobile network goes down. On the shop floor, no one notices: the machines are running, the wired network is working. Only a few operators notice their phones display "SOS only."
11:45 AM: a machining center enters a lubrication fault state. The alarm is triggered, as planned, via SMS to the on-call technician, who is out for lunch. The SMS is never sent. The supervisor doesn't know: from their perspective, the alert was sent.
12:30 PM: the logistics manager tries to reach the storage site, connected only by a 4G router. No phone, no messaging, and the remote site's inventory is no longer visible in the ERP. Order picking is put on hold.
1:15 PM: the team leader discovers the halted machine upon returning to their station. An hour and a half of lost production, plus a batch of parts to inspect to ensure none were produced out of tolerance before the shutdown.
2:00 PM: a press displays a fault that the in-house team cannot diagnose. The manufacturer usually intervenes remotely via a remote maintenance box equipped with a SIM card. Unable to connect. The manufacturer's technician won't be able to travel on-site until the next day.
4:30 PM: management realizes that the temperature sensors in the storage site's cold room, also running on the mobile network, haven't reported anything for five hours. Someone takes their car to go check on-site.
8:30 PM: the network comes back, and queued SMS messages arrive all at once, including the 11:45 AM alarm, now useless.
The next morning: the assessment sums up in just a few lines: a machine stopped for an hour and a half, a press immobilized until the technician arrives, shipments delayed by a day, and five hours without any data from a remote site. No equipment in the plant actually broke down. It was an invisible dependency that failed.
There is nothing exceptional about this scenario! Each of these situations stemmed from a reasonable choice made in isolation at the time it was decided. It is their accumulation on a single carrier, which no one had ever mapped out, that turns an external outage into a lost day.
Three lessons for your industrial systems
1. Map your single points of failure
A core network is the engine that everyone shares. When it stalls, all cell towers across the country become useless at the same time. Your IT/OT system inevitably has its equivalent: a single carrier, a license server, a VPN gateway, a directory. The first step is to list them, flow by flow, and note for each one what grinds to a halt downstream.
2. Treat every migration as a high-risk operation
If the press hypothesis is confirmed, the outage didn't stem from an external incident, but from a planned change. This is the most common scenario in industry as well: version upgrades, infrastructure changes, new firewalls. A well-executed migration includes a phased rollout, a tested rollback plan, and a maintenance window chosen based on production schedules, not the IT calendar.
3. Design for degraded mode
A factory must not stop just because the Internet stops. Critical processing benefits from running as close to the machines as possible, with the ability to store locally and resynchronize later. A secondary communication path, using another carrier or a different technology, turns a blocking outage into a minor slowdown.
How we approach the topic at akawan
We don't sell network services, and no one can promise zero downtime. Our job is to ensure that an external cutoff doesn't turn into a production halt.
Kaptngo, our secure communication gateway, was designed for this type of context: replacing aging firewalls and VPNs, handling machine-to-machine data flows, legacy protocols, or license servers, and executing processing as close to the data as possible, including in isolated environments (Kaptngo project).
We also address the issue upstream on the monitoring side: reporting anomalies reliably when connectivity is constrained (Optimized anomaly reporting). And on the infrastructure side, migration is part of our daily work, with the same standard of continuity that was missing on October 21 (Infrastructure migration).
To explore field constraints in greater depth, our article Tech in Industry: Facing the Challenges of Disconnected Environments details why a factory cannot think like a permanently connected office.

Une journée et demie sans Outlook
Vos courriels sont revenus, pas vos décisions…

When the automaton lies...
who sounds the alarm?

The AI agent that clicks for you
… and takes your data where no one is looking

The demonstrator who has never seen a production rate
If your AI pilots are not making it to production, the problem is not the AI

How many open houses in your workshop?
No one in the company knows the exact number…

The digital health record you don't have
The digital health record that your industry has never opened.

Multi-sites, zero visibility...
The nightmare of blind industrial management.

Industrial M&A: The Billion-Dollar Bug
IT architecture secretly dictates the success of your mergers

Successful merger, ERP in ruins
The true price of your redemptions…

Why 80% of factory mergers verge on a digital crash
and how to escape it…



