About
Subscribe

Why failover testing matters

Johannesburg, 20 Nov 2026
Why failover testing matters.
Why failover testing matters.

On 12 November 2026, Microsoft 365 began a global outage that ran for just over 30 hours. Exchange Online was hit first. Mailbox access through Outlook on the web, mobile sync and the REST API stalled, and mail moving between corporate mail systems and Exchange Online backed up. Teams, SharePoint Online and Microsoft 365 Copilot rely on Exchange for parts of their functionality, so they degraded with it. Microsoft closed the incident on 13 November and published its explanation on 16 November, with a final post-incident report to follow.

The cause is the most instructive part of the story. Microsoft says the outage was caused by a latent bug introduced during an earlier network capacity upgrade, which sat undetected until routine maintenance rerouted some traffic down a different path. Once that path carried real load, Exchange Online connection requests surged and piled up in queues, and the affected servers hit their processing limits and stopped accepting new connections. Microsoft has stated that the incident was a technical fault with no link to a cyber attack. Recovery took longer than the fix itself because demand stayed high, so services were brought back in stages, with stability confirmed at each step.

So, when did your organisation last test its failover?

A failover is a reroute. It moves your workloads, data and users onto a path that spends most of its life idle. The flaw lived in exactly that kind of path, one that had been built and upgraded but had never carried that load until the day it mattered. Every recovery environment that has gone untested holds the same kind of latent assumptions. Replication drifts out of step with production. Application dependencies change as systems are added. Runbooks age as teams and infrastructure move on.

The larger the business, the more this matters. Scale creates interdependence, and interdependence turns a single fault into a chain reaction. Microsoft's incident showed this, where one service hit its limits and the services built on it followed. In a large enterprise, a core system that fails to come up at the recovery site can stall finance, operations, customer channels, supply chain and every other major department at once. The cost of that delay shows up in lost revenue and the trust of customers who expect continuity as a given.

Testing is how an organisation finds out what its recovery looks like. A tested failover tells you how long recovery takes in practice, which dependencies need to come up first, who does what and whether the data on the other side is complete and usable. It turns a recovery time objective from a figure in a policy document into a number the business has seen with its own eyes. It also gives leadership something they can stand behind.

Microsoft's staged, verified recovery is worth noting for the same reason. Bringing services back in a controlled sequence, checking stability at each step, is what a rehearsed recovery looks like. That discipline comes from practice, and practice comes from testing regularly, after significant changes, and under conditions close enough to the real thing to be meaningful. Microsoft's own account makes the point: its normal load testing missed the bug because the traffic mix in its test labs differed from what the live service carried.

At Resilient Innovations, testing plays a crucial role in how recovery is approached. Its Disaster Recovery as a Service, built with Veeam on Resilient Innovations' local private cloud, allows failover to be exercised in isolated environments without disrupting production, so businesses can prove their recovery as often as their risk profile demands. Through Resilient Innovations' Risk Advisory Services, its Business Continuity Management practice aligns that testing with what the business needs to keep running, so the systems that matter most are the ones proven first. That includes the platforms you do not run yourself. Mapping your reliance on services like Microsoft 365, and planning how your people work and communicate when one goes down, belongs in the same exercise. With close to four decades in business continuity and disaster recovery, Resilient Innovations has seen that the organisations that recover well are the ones that have already done it before the event arrived.

Share