Summary

  • Rogers says a March 4, 2026 exercise simulated a large-scale core outage for four hours, with test devices remaining connected and operational.
  • That is more than a promise. It is also a company account with important details withheld, not independent certification that every future outage has been prevented.

For four hours on March 4, 2026, Rogers put a disaster-recovery system through an exercise simulating a large-scale failure of its core network. According to its July filing to CRTC, the transition caused no service disruption: all test devices stayed attached, authenticated and operational, and performance met the expected thresholds.

The word “test” matters. Those were devices in the exercise, not a demonstrated result for every customer in Canada. The public filing withholds traffic figures and much of the description of the test environment. But the date and the four-hour duration are there. It is no longer accurate to describe the public record as a network-separation promise with no dated test account behind it.

For Rogers, the significance goes back to a failure that left even the people fixing the network struggling to reach it.

The repair team needed another way in

On July 8, 2022, an outage affected more than 12 million wireless and wireline customers. Payments and access to emergency services were caught in the disruption. The independent Xona assessment traced the trigger to the sixth phase of a seven-phase upgrade. Removing a routing-policy filter let too much routing information reach core routers, overwhelming them without the necessary overload protection.

Earlier phases had gone well. That helped turn a high-risk change into a low-risk one in the approval process, removing higher scrutiny and laboratory testing. A successful run of changes had become a reason to look less closely at the next.

Then the recovery effort hit its own dependency. Management access used the network that had failed. Critical sites lacked connectivity from another provider. Staff had to reach equipment physically and obtain other carriers' SIM cards to communicate. This was not merely a shortage of spare network capacity; the route used to repair the failure was also broken.

Xona subsequently recorded safeguards on routers, separate management connectivity, third-party backup communications and improved procedures. Its overall assessment was positive. It did not find that sharing a core was inherently a design flaw, although the shared dependency broadened the outage's effects.

Three different kinds of separation

Keeping engineers connected is one job. Separating the customer-facing wireless and wireline data cores is another. Providing a route for service to recover when a core fails is a third. The full Xona report, made public in November 2024, discusses these distinctions. Two cores do not automatically protect each other from the same bad change, and a backup arrangement brings its own capacity and operating requirements.

The regulatory record also moved beyond the first announcement. In July 2024, CRTC said the measures had addressed the cause identified by Xona, while requiring a further report on effectiveness and the core split. Rogers' July 2025 response marked 16 recommendations as in effect and effective. Those entries are company attestations, not 16 independently witnessed tests.

A December 2025 letter requested another update. The resulting July 2026 filing says Rogers developed and implemented a new disaster-recovery solution during 2025, then exercised it the following March. Much of the architecture is redacted; the disclosed test cannot simply be relabelled proof that the precise physical split promised earlier is complete.

The useful change is narrower and real: the record now contains a described recovery exercise and reported results. Xona's earlier independent assessment did not witness that later exercise. Nor does the public test account identify an independent observer or disclose enough to reconstruct its full scope.

For customers, a successful rehearsal is welcome. To judge how much confidence it earns, the next useful details are the load it carried, the failures it covered and the exceptions it found. Four hours can demonstrate something important without guaranteeing everything.

Historical references: the original Canadian government announcement and CBC coverage remain linked. The subsequent findings above are supported by the regulator's reports and company filings, checked on October 19, 2026.