Why My VPN Kept Disconnecting on Phone and Laptop—and the Connection That Finally Stayed With Me

The client’s face froze halfway through a sentence.

My laptop showed that the VPN had disconnected. A second later, it reconnected and the meeting returned—but the presentation I had been sharing did not. I apologised, opened the file again and continued. Three minutes later, the same thing happened.

I was staying with family on Dongyin, one of Taiwan’s Matsu Islands. The guesthouse Wi-Fi had been uneven all morning, so I moved the laptop to my phone’s hotspot. The VPN connected, held for less than a minute and dropped again. I blamed the laptop until the VPN icon began cycling on the phone too.

The local connection had a reason to be temperamental. In April 2026, damage to a Taiwan–Matsu subsea cable forced authorities to move Dongyin’s voice and internet traffic onto backup microwave transmission. Basic service remained available, but Taiwan’s Ministry of Digital Affairs warned that some internet services could experience delays. (Gov)

That was exactly what the network felt like. It was not fully offline. Messages arrived. Websites opened. Speed tests occasionally looked normal.

But the path beneath the VPN kept weakening or changing.

I had 22 minutes before presenting the final version of a campaign plan. I did not need an impressive speed-test result. I needed the meeting to survive six more slides.

Reconnecting Was Not the Same as Recovering

My established VPN appeared to be fixing itself.

Whenever the underlying connection dropped, the app noticed and reconnected automatically. Its status changed from green to grey, spun for several seconds and turned green again.

The problem was what happened during those seconds.

The meeting lost audio. Screen sharing stopped. The cloud presentation sometimes returned to its sign-in page. A file upload restarted from the beginning. On the phone, messages remained stuck until I opened the VPN app and watched it reconnect.

The tunnel came back. The task did not.

That distinction matters because a VPN sits on top of another connection. When a phone moves from Wi-Fi to mobile data, or a laptop leaves a router and joins a hotspot, the old route disappears. The VPN must follow the new route quickly enough that the meeting, upload or message session does not give up first. (Google)

A green icon returning after ten seconds may look like successful recovery.

Inside a live meeting, ten seconds is a disconnection.

Once I saw the difference, the repeated loop stopped looking like a minor inconvenience. The VPN was not merely reconnecting too often. It was making every brief network change visible to everything I was trying to do.

I Kept Repairing the Devices

I started with the laptop.

I restarted it, forgot the guesthouse Wi-Fi and joined again. I switched the VPN from automatic mode to another protocol. I chose a nearby server, assuming the shorter distance would make the connection steadier.

The next interruption arrived during the same slide.

Then I moved to the phone. I disabled battery-saving mode, restarted the hotspot and confirmed that the VPN could continue running in the background.

For a few minutes, everything seemed calmer. Then the Wi-Fi weakened, the phone changed network paths, and the disconnect–reconnect loop returned.

Testing another network is useful because it separates a device problem from a connection problem. (Apple) I had now tested two devices, the guesthouse Wi-Fi and mobile data. Without the original VPN, both devices regained ordinary internet access quickly. With it, the same interruptions followed me from one network to the next.

A recent public discussion described the same practical frustration: repeated VPN reconnects across two laptops and several connection types, even after changing protocols. (Reddit) The useful point was simple.

That changed my diagnosis.

The established provider was not generally bad. It had a mature network, extensive documentation and more server choices than I would ever use. On stable home broadband, it had worked quietly for months.

Its weakness appeared when the connection underneath it changed.

Speed Was Measuring the Wrong Moment

I ran another speed test while the VPN was connected.

The result looked good. Download speed was sufficient for video, and the latency number did not seem alarming. For a moment, I thought the latest restart had worked.

Then the Wi-Fi disappeared for two seconds.

The VPN reconnected. The speed-test page recovered. The meeting did not.

That was when the comparison became clear. Speed described the network while everything was healthy. It said very little about what happened during the interruption.

On stable broadband, maximum throughput may decide whether a large file takes four minutes or five. On weak Wi-Fi or mobile data, recovery decides whether the file finishes at all.

The provider’s server list did not help either. Moving between Taiwan, Japan and Singapore changed the destination, but every tunnel still had to survive the same unstable connection underneath it.

I had twelve minutes left.

Instead of choosing another location, I opened the smaller backup already installed on my phone.

The Meeting Returned—and Stayed

I launched OnlydogVPN and selected its preset for a weak or changing network.

There was no country list to work through and no conventional protocol decision to make. I connected the laptop through the guesthouse Wi-Fi, reopened the presentation and joined the client again.

I advanced to the next slide.

A minute later, the Wi-Fi weakened. The laptop lost its route through the guesthouse network and moved to the phone hotspot I had left available.

The video softened.

It did not disappear.

The audio continued. The presentation remained on screen. I finished explaining the revised launch timeline without asking anyone to wait.

Then I watched the file upload that had already failed twice. It paused briefly when the network changed, resumed and completed.

That was the result I had needed from the beginning. The meeting stayed alive, and the file reached the client. I did not have to reopen either one.

The service uses HTTP/3-based transport and is designed to recover across weak or changing networks. HTTP/3 runs over QUIC, which can preserve a connection when the device’s network path changes instead of binding the entire session to the route on which it began. (IETF)

The important part was not the protocol name. It was what happened on the screen.

Wi-Fi weakened. The hotspot took over. The work continued.

I could not inspect the service’s internal routing and recovery rules, but the visible difference was consistent throughout the test: the original VPN repeatedly rebuilt its tunnel after a change, while the smaller app carried the active session across it.

That was the fix—not preventing every network drop, but preventing every drop from destroying the task above it.

The Phone Had Been Part of the Same Problem

After the meeting, I needed to send the client a revised PDF from my phone.

With the first VPN, the phone often became stuck after moving between the guesthouse Wi-Fi and mobile data. The VPN icon remained visible, but messages stopped sending until I disconnected manually.

The smaller service let me connect the phone through a verification code shared from the laptop, without entering another email address and password.

The PDF sent.

I walked downstairs, moved out of Wi-Fi range and continued replying over mobile data. The conversation stayed active through the switch.

That smaller result gave me a reason to keep the app on both devices. My working connection was not one laptop attached to one router. It was a laptop using a phone hotspot, a phone moving between Wi-Fi and mobile data, and several cloud tools that reacted badly to interruptions.

Treating each device as a separate troubleshooting project had made me repeat the same work twice. Connecting both through a service built around recovery made the setup behave more like one continuous workspace.

When the Usual Fixes Are Still Worth Trying

Some reconnect loops do begin with the device or local network.

If every device loses internet at the same time, the router or internet provider may be down. If only one laptop disconnects from every Wi-Fi network, its network settings or drivers deserve attention. If the problem began after an operating-system update, updating the VPN app may restore compatibility.

Battery-saving settings can also suspend a VPN in the background on a phone. A kill switch can make a brief tunnel interruption look like a complete internet outage because it blocks unprotected traffic until the VPN returns.

Those checks are useful because they narrow the cause.

They should not become an endless ritual.

In my case, ordinary internet traffic returned quickly. The reconnect loop followed the original VPN across two devices and multiple network types. Changing protocols and server locations did not stop it.

Once that pattern was clear, restarting everything again would only have consumed the rest of the meeting.

Recovery Mattered More Than Stability

I had spent the morning trying to make the underlying connection stable.

On Dongyin that week, stability was not entirely mine to control. Regional traffic had been moved to backup infrastructure. The guesthouse Wi-Fi weakened as more people used it. Mobile signal changed from one room to another.

No VPN could repair the subsea cable or strengthen the local signal.

What it could do was handle those changes without turning each one into a fresh crisis.

The smaller service has fewer server locations, a shorter public history and fewer independent reviews than the major provider I tried first. Those differences matter when broad geographic choice is the main requirement.

They did not matter during my presentation.

The established provider was fast whenever the connection held. The backup was better at the moment that decided whether I could continue working: when Wi-Fi weakened, the hotspot took over and the VPN had to follow.

By the end of the call, I had stopped asking how to prevent a VPN from ever disconnecting.

On an unstable network, the connection that matters is the one that lets the work continue before anyone notices it had to recover.