Why Train Wi-Fi Keeps Dropping Your VPN Even When the Signal Looks Full

The train had barely cleared the station when the publishing page stopped responding. My VPN icon turned yellow, the browser went offline, and the client’s final price update remained one click away from going live. I blamed the train Wi-Fi, disconnected, joined again, and watched the signal return to full strength. Then the VPN began reconnecting for the second time.

The update had to be published before ten. The client’s campaign was already scheduled, and several people were waiting for the new page to appear.

Without the VPN, the publishing dashboard loaded. With it connected, I could work for a minute or two before the tunnel dropped and the page lost its session.

This was my own travel VPN, not an employer-managed corporate tunnel, so I was free to change it. What I could not change was the train, the route, or the deadline.

I restarted the VPN and selected another server.

The connection returned just as the train entered a cutting.

It disappeared again before I could press Publish.

Full Wi-Fi Bars Were Telling Me Only Half the Story

Train Wi-Fi looks familiar because my laptop connects to a router inside the carriage, much as it would at home. But the carriage router still has to reach the internet from a train moving through tunnels, valleys, stations, and crowded mobile cells.

Rail Wi-Fi systems commonly use roof-mounted antennas and several mobile links, passing whatever connection is available to passengers inside the train.

That creates two separate connections:

My laptop was connected to the carriage.

The carriage was trying to stay connected to the outside world.

The first connection could show full bars while the second was switching carriers, losing coverage, or sharing limited capacity with hundreds of passengers. The Wi-Fi icon only confirmed that my seat still had a strong link to the train.

Recent UK measurements show how unreliable the link beyond the carriage can be. Ofcom found poor mobile performance in 58 to 83 percent of its train tests, depending on the network. Government plans to improve trackside infrastructure and address dozens of tunnel blackspots show that this is a structural rail-connectivity problem, not simply an unlucky laptop.

Once I understood that, the repeated VPN drops made more sense. The network beneath the tunnel kept changing as the train moved.

The next question was whether my trusted VPN could adapt to that movement.

The Established Provider Looked Strongest at the Station

The major provider was the reasonable first choice. It had a long public history, a large support operation, and servers in more locations than I would ever need.

At the station, it had performed perfectly.

I ran a speed test while waiting on the platform. The result looked fast enough for video, let alone editing a mostly text-based webpage. I selected a nearby server because the shorter route seemed likely to be the most stable.

Once the train accelerated, that advantage disappeared.

The tunnel connected quickly, but every interruption forced it to start again. Its kill switch blocked ordinary traffic while the VPN rebuilt the connection. That protected the session from continuing outside the tunnel, but it also turned every brief coverage gap into a complete stop.

I changed servers after the first drop.

I changed protocols after the second.

One combination lasted nearly five minutes, long enough to make me think I had fixed it. Then the train crossed another weak section and the publishing dashboard returned to its login screen.

The provider was fast whenever the path stayed stable. The path simply did not stay stable.

Other commuters have described the same practical frustration: onboard Wi-Fi can be usable, yet work VPNs still require repeated reconnections during the journey. That was enough to confirm the pattern. A train connection did not have to remain offline for long to destroy an active session.

By then, changing another server felt less like troubleshooting and more like resetting a stopwatch.

My Phone Hotspot Removed One Layer—and One Advantage

I switched off the onboard Wi-Fi and connected the laptop to my phone.

For several minutes, this looked like the answer. The VPN connected, the dashboard opened, and my edits were still there.

Then the train entered a tunnel.

The phone lost mobile service. The VPN dropped, the browser froze, and the publishing system asked me to sign in again when the signal returned.

The hotspot had removed the train’s shared Wi-Fi, captive portal, and onboard router. It had also replaced the train’s roof antennas and multiple network links with one phone on one carrier inside the carriage.

I moved the phone toward the window and tried again. The signal improved, but the next weak section produced the same result.

At that point, choosing between train Wi-Fi and a hotspot meant choosing which interruption I preferred.

That changed the problem completely. I was no longer looking for a network that would never drop. On a moving train, that was unrealistic.

I needed a VPN that made a short drop less destructive.

The Smaller App Treated Movement as Normal

I opened OnlydogVPN and selected its preset for weak or changing networks.

There was no map to study and no list of cities to test one by one. I connected, reopened the publishing dashboard, and made the final two edits.

The train Wi-Fi weakened again.

The page paused.

This time, the VPN did not fall into a visible reconnect loop. When the train recovered its external connection, the dashboard continued from the same session. The unsaved text remained in place, and I did not return to the login screen.

I clicked Publish.

The confirmation appeared before the next station.

That result mattered more than another speed test. The connection had still crossed a weak section. The difference was that the interruption no longer erased the task.

The service uses an HTTP/3-based transport designed to recover across changing network paths. Put simply, it is better suited to keeping a session alive when the route beneath it shifts between available connections, instead of rebuilding everything from the beginning.

That matched what was happening on the train. Even though my laptop remained connected to the same carriage Wi-Fi, the train’s external route kept changing as coverage came and went.

I could not observe the operator’s internal routing, filtering, or carrier-selection rules, so I could not identify the precise trigger behind every earlier disconnect. The visible result was clear: the established service repeatedly rebuilt its tunnel and returned the dashboard to its login page; the smaller app preserved the session long enough to publish the update.

Its interface also stopped me from making the problem worse.

Every server switch had created another experiment. Every protocol change had required another wait. The next tunnel or coverage gap often arrived before I had learned anything useful from the previous test.

The smaller app asked about the situation rather than the geography. Once I chose the weak-network preset, I could return to the work instead of managing the VPN.

The Page After Publishing Revealed a Smaller Advantage

After the update went live, I opened the client’s public homepage to confirm that the new pricing appeared correctly.

The page loaded, and the app’s blocked-request counter began rising. Advertising and tracking requests were being stopped before they competed for the train’s limited connection.

That did not strengthen the mobile signal outside the carriage. It simply reduced the amount of unnecessary traffic using the connection that remained.

On home broadband, a collection of background requests is easy to ignore. On shared train Wi-Fi, each extra request becomes more noticeable. The page I needed appeared, the new price was correct, and fewer unrelated connections were fighting for the same narrow route.

It was a small discovery after the main task had succeeded, but it gave me a reason to keep the app installed after the journey.

The Trade-Off Was Still Real

The smaller service has fewer server locations than the established provider. It also has a shorter public history and fewer independent reviews.

Those are meaningful limitations. A larger provider’s track record, infrastructure, and support operation deserve weight when choosing a privacy product for long-term use.

They simply did not decide what happened on this train.

The established provider performed well while the network path stayed stable. My phone hotspot avoided the onboard Wi-Fi but depended on one carrier signal inside a moving carriage. Both lost the publishing session when the connection beneath them disappeared.

The smaller app accepted that the route would change and recovered without destroying the work already in progress.

That was the comparison I had missed at the station.

Train Wi-Fi does not keep disconnecting a VPN because the Wi-Fi icon is lying. The icon describes the strong link between your device and the carriage while saying almost nothing about the train’s changing connection beyond it.

On that journey, the useful VPN was not the one that looked fastest before departure. It was the one that let me press Publish once.