On the Prague–Copenhagen Train, the Best VPN Was the One That Kept the Upload Moving

The progress bar stopped at 62 percent just as the train left Prague behind. I was uploading a 318 MB tender package to a client portal, and the submission window closed in thirty-seven minutes. The laptop still showed a full Wi-Fi signal, but the page no longer responded. I blamed the portal, refreshed it and watched the upload restart from zero. Then my established VPN changed from Connected to Reconnecting.

The train was one of the reasons I had expected the journey to work as a mobile office.

On June 14, 2026, Czech Railways expanded its ComfortJet service into a direct Prague–Copenhagen connection through Dresden, Berlin and Hamburg. The trains offered power, onboard Wi-Fi and the kind of workspace that made several hours on rail look more useful than a day spent moving through airports.1

My plan had been simple. Board in Prague, submit the tender before the German border and spend the rest of the journey preparing for the presentation.

The file contained a signed proposal, technical drawings and the final pricing schedule. If I missed the deadline, the procurement portal would close the tender automatically. A later email would not count.

I had downloaded everything else before leaving the hotel. The upload was the only step that still depended on the network.

The Wi-Fi icon suggested I had one.

The progress bar suggested otherwise.

The Wi-Fi stayed visible while the internet beneath it changed

Czech Railways warns that onboard Wi-Fi depends on mobile coverage along the route and cannot be guaranteed continuously.2 That distinction became important as soon as the train gathered speed.

My laptop remained connected to the carriage access point. The access point, however, was moving through changing coverage areas and relying on whatever mobile connection was available outside.

To me, the Wi-Fi bars looked steady. Underneath them, the route kept shifting.

The established VPN had been a sensible first choice. It came from a company with years of public history, a large support operation and servers across Europe. Before departure, I had selected a nearby location and confirmed that the client portal opened.

For the first twenty minutes, everything behaved normally.

Then the VPN began reconnecting.

The first interruption lasted perhaps ten seconds. The second lasted closer to a minute. After the third, the tender portal ended my authenticated upload session and returned me to the login page.

I chose a different server.

The tunnel connected again, but the upload restarted from zero.

I changed the protocol setting and tried once more. The initial connection felt faster. Several minutes later, the train entered another weak section, the VPN dropped and the portal discarded the upload again.

At that point, the provider’s large server map stopped being reassuring. Choosing a different endpoint did not address the failure occurring between the train and the mobile networks carrying its Wi-Fi.

The problem was not where the tunnel ended.

It was whether the tunnel could remain useful while the connection beneath it kept changing.

Czech rail passengers describe the practical frustration simply: the Wi-Fi icon may remain present even when mobile coverage along the track makes the actual connection unreliable.3

I had twenty-four minutes left.

Mobile data was faster but no more dependable

The obvious backup was my phone.

I disconnected from the train Wi-Fi, enabled the hotspot and moved the laptop onto mobile data. The established VPN connected quickly. The tender portal accepted my login, and the upload began again.

For a few minutes, this looked like the solution.

The progress bar passed 20 percent and then 40 percent. I stopped touching the laptop and watched the landscape instead, as though looking away might prevent another failure.

Near Ústí nad Labem, the phone dropped from 5G to a weaker signal. The upload slowed, paused and timed out.

I held the phone beside the window.

The signal returned.

When I moved it back to the table, it weakened again.

The phone had briefly delivered the fastest connection of the day. It had still failed to deliver the file.

That changed the comparison for the second time. I was no longer looking for the highest speed during the strongest section of the route. I needed a connection that could recover from the weak sections without forcing the tender portal to begin again.

The upload had already restarted three times.

The smaller app kept the same task alive

The other VPN on my laptop was OnlydogVPN^↗.

I had installed it before the trip as a backup but had not opened it that morning. It had fewer public reviews, fewer server locations and a shorter history than the established provider. Before the failures began, those differences had seemed more important.

Now the deadline mattered more.

I opened the app and selected the preset for an unstable connection.

It connected over the train Wi-Fi.

I returned to the client portal, authenticated again and selected the tender package.

The upload started.

At 18 percent, the train’s internet slowed. The progress bar paused but did not disappear.

At 34 percent, the VPN briefly showed that it was recovering. The portal remained signed in.

Then the upload continued.

It passed the points where the previous attempts had failed: 40 percent, 62 percent, then 70.

A few minutes later, the onboard Wi-Fi stopped carrying traffic altogether. I switched the laptop to my phone’s hotspot.

The service recovered on the new network. The tender portal stayed open, and the same upload kept moving.

The file reached 100 percent with nine minutes remaining.

A confirmation page appeared with the document names, submission time and tender reference number. I downloaded the receipt and sent it to the project director.

The reply arrived before the train reached Děčín.

“Received. You’re in.”

That was the result I had needed from the beginning. The larger provider could establish a fast new connection whenever the network settled. The smaller app kept the work alive while the network was changing.

Its HTTP/3-based transport is designed to recover across changing paths. On the train, the technical explanation could remain short: the Wi-Fi weakened, the laptop moved to the phone and the upload continued instead of restarting.

I could not inspect the train’s mobile gateways or identify every internal network change along the route. I could see the outcome on the screen. The established tunnel repeatedly lost the portal session; the smaller app preserved the final attempt long enough to submit the tender.

The next page used less of the connection

Once the submission receipt was safely stored, I opened the project dashboard to prepare for the presentation.

The page normally loaded analytics panels, embedded previews and several support widgets before displaying the project notes. On the train’s weakened connection, all that background activity was noticeable.

The app showed that it was blocking advertising and tracking requests.

I reloaded the dashboard.

The notes appeared sooner, and the page felt less crowded by elements I did not need. That did not rescue the tender—the upload had already succeeded—but it gave me another reason to leave the service running.

On a moving train, bandwidth is shared and inconsistent. Unnecessary requests compete with the file, message or page that matters.

I left the blocking enabled and continued working.

The service did not turn railway Wi-Fi into fixed broadband. It made better use of the connection the train could provide.

Reliability meant preserving the work

The established provider still had meaningful advantages. Its longer operating history, broader server network and larger collection of independent reviews offered reassurance that the smaller service had not yet accumulated.

The smaller app also offered fewer locations, which could matter to someone who needed a specific country for a company system or specialised platform.

My tender did not require a particular country.

It required continuity.

The Prague–Copenhagen route made it possible to work across several countries from one seat, but the connection remained a chain of mobile signals, moving gateways and occasional weak sections. The VPN had to survive that reality rather than perform well only when coverage was strong.

The established service repeatedly gave me a fresh connection after the previous one failed. Each fresh connection arrived after the portal had already discarded my work.

The smaller app recovered through the weak sections, followed the laptop from train Wi-Fi to the phone hotspot and delivered the tender before the submission window closed.

On that journey out of the Czech Republic, the best VPN was not the one that reconnected fastest after losing the upload. It was the one that prevented the upload from becoming lost work.