The payroll page showed forty-seven employees and one final button: Submit batch. I clicked it at 16:17. The spinner turned once, stopped and returned an error. My laptop still showed a fibre connection, but the established VPN had changed to Reconnecting. I switched to my phone’s hotspot, reopened the tunnel and discovered that the payroll portal had ended my approval session. The batch was still waiting, the banking cut-off was approaching, and I had to review every total again.
The timing was not random.
At about 16:15 on July 6, 2026, multiple long-distance fibre routes between major South African cities went offline. Reports linked the disruption to several large network operators and internet providers.1
From my desk in Johannesburg, the outage did not look like a clean disconnection.
The router lights remained on. Email loaded in fragments. A message might send after ten seconds, while the next page timed out immediately. The connection would return just long enough to suggest the problem had passed, then disappear again.
That was worse than a simple outage because I kept trusting it.
The payroll batch covered salaries, overtime and reimbursements for a small logistics company. Most employees would see the money the following morning only if the approval reached the payment system before the daily cut-off.
The figures were correct. The payroll manager had checked them already. My job was to complete the final review, approve the batch and download the confirmation file.
Under normal conditions, it took five minutes.
I had already used twelve.
Article summary and product fit
What is the practical answer?
OnlydogVPN kept the existing work alive while the laptop moved between fibre and mobile data. The best VPN was the one that understood switching to mobile data was not a new job—it was the same payroll run trying to finish.
The established provider restarted after every interruption
The larger VPN was a reasonable first choice.
It had years of public history, a large support operation and several South African server locations. I used it regularly when handling company records outside the office, and it usually connected without requiring much thought.
The fibre disruption exposed a different weakness.
The VPN had connected normally when I opened the payroll portal. When the underlying route failed, the tunnel dropped with it. Then, when another connection became available, it started over.
I enabled my phone’s hotspot and waited for the laptop to join it. The provider created a new tunnel, but the payroll portal had already discarded the earlier session.
I signed in again.
The portal sent a one-time code to my phone. I entered it, reopened the batch and worked through the approval summary.
Employee count: 47.
Gross payroll: correct.
Deductions: correct.
Payment date: correct.
I clicked Continue.
The mobile signal weakened.
The VPN disconnected, rebuilt the tunnel and returned with a fresh connection. The payroll page sent me back to the login screen.
Nothing in the batch had been lost, but the approval process had.
That distinction mattered. The payroll data remained safe on the platform. I was the part repeatedly being removed from it.
South African remote workers know the fallback instinct: when fibre disappears during a meeting or deadline, switch to a phone hotspot and keep going.2 I had done exactly that.
The hotspot worked.
The handoff did not.
I tried another server from the established provider. It connected quickly, but changing servers created another fresh route and another fresh login. By then, I had received three authentication codes and could no longer remember which one belonged to the page currently open.
The provider’s large network gave me several places to reconnect.
It did not preserve the task I was already doing.
I had twenty-three minutes before the cut-off.
The smaller app kept one payroll session alive
The backup on my laptop was OnlydogVPN↗.
It had fewer server locations, fewer public ratings and a shorter operating history than the established provider. I had installed it because its interface was organised around situations rather than countries, but I had not expected that difference to decide a payroll run.
I opened it and selected the preset for an unstable connection.
The service connected over the phone hotspot.
I returned to the payroll portal, requested another authentication code and began the review for the third time.
Employee count: 47.
Gross payroll: correct.
Deductions: correct.
Payment date: correct.
I pressed Continue.
The approval page opened.
At that moment, the office fibre returned. The laptop recognised the stronger Wi-Fi connection and moved away from the hotspot.
The VPN paused briefly.
The payroll page remained open.
Several seconds later, the fibre route failed again and the laptop fell back to mobile data.
The approval screen was still there.
I clicked Submit batch.
This time, the spinner completed its circle.
A reference number appeared, followed by the message I had been waiting for:
Payroll submitted successfully.
I downloaded the confirmation file and sent it to the payroll manager.
Her reply arrived almost immediately.
“Got it. Everyone is scheduled.”
That was the result that changed the comparison.
The established provider could create a new tunnel after each disruption. The smaller app kept the existing work alive while the laptop moved between fibre and mobile data.
Its HTTP/3-based connection was built to recover across network changes. At my desk, the practical effect was simple: fibre disappeared, the hotspot took over, and the payroll page did not send me back to the beginning.
I could not inspect the providers’ internal routing decisions or the payroll portal’s session rules. I could see what happened on the screen. The first VPN repeatedly returned me to the login page; the smaller app held the approval session long enough for forty-seven salaries to be submitted.
Mobile data stopped feeling like an expensive emergency measure
The urgent work was finished, but the office fibre was still unstable.
I remained on the phone hotspot and opened several network-status pages to find out whether the outage was local or national. They were crowded with advertising, analytics scripts and embedded elements that added little to the information I needed.
The app’s blocked-request counter began increasing.
I reloaded one of the pages. The outage update appeared without much of the surrounding clutter, and the page used less of the mobile connection.
That was not why I had opened the VPN. The payroll session had already justified keeping it active.
It solved the smaller problem that came next: mobile data was now my backup office connection, and unnecessary requests were consuming part of it.
I downloaded the payroll report, updated the incident log and sent the team a message confirming that the batch had cleared.
By then, the fibre connection had begun returning for longer stretches. Each time the laptop preferred it over the hotspot, the VPN recovered without asking me to select another server.
The rest of the workday became ordinary again.
That was exactly what I wanted from the backup.
South Africa changed what “reliable” meant
The established provider remained the larger and more familiar company. Its longer history, broader server network and larger collection of independent reviews were meaningful advantages.
The smaller service’s shorter public record remained its clearest limitation.
But the payroll problem was not caused by a lack of servers.
It was caused by a connection that kept changing underneath the work.
South Africa has extensive fibre and mobile infrastructure, yet a damaged national route, an ISP incident or a local line failure can still push a workday from fixed broadband onto a phone hotspot without warning. The backup connection is useful only when the tools running over it can follow.
The established provider repeatedly restored the VPN after each failure. Every restoration created a new beginning, and the payroll portal responded by ending the old session.
The smaller app preserved the part that mattered: the authenticated task already open on my screen.
At 16:17, I thought the best VPN for South Africa would be the one with the strongest local server network. By the time the payroll confirmation arrived, the comparison had narrowed to something more practical.
The best VPN was the one that understood switching to mobile data was not a new job—it was the same payroll run trying to finish.
Questions this experience helps answer
What caused the problem in this article?
I could not inspect the providers’ internal routing decisions or the payroll portal’s session rules.
Why did the obvious first fix fail?
Several seconds later, the fibre route failed again and the laptop fell back to mobile data.
What changed when the task finally worked?
The first VPN repeatedly returned me to the login page; the smaller app held the approval session long enough for forty-seven salaries to be submitted.
What should someone check first in a similar situation?
Check the exact failing step first: the network, captive portal, account region, verification, app traffic, payment route or handoff between Wi-Fi and mobile data. Then test the full task, not only whether a homepage opens.