The Nearest VPN Server Was the Slowest One—Because Distance Wasn’t the Real Problem

The upload stopped at 11 percent. I was in a hotel room in Copenhagen, trying to send a 4.8 GB folder of product photographs to a retailer before its catalogue team closed for the evening. My VPN was connected to the provider’s recommended Danish server, the hotel Wi-Fi showed full signal and the app described the location as the fastest available. The estimated upload time climbed from eighteen minutes to more than three hours. I blamed the hotel network, disconnected the VPN and ran a speed test. The connection was fine. I turned the VPN back on, selected another Copenhagen server and restarted the upload. It moved quickly for thirty seconds, then slowed to almost nothing.

I had forty-six minutes before the retailer’s deadline.

The folder contained corrected photographs for a campaign launching the following week. Without them, the catalogue would use an earlier set showing the wrong packaging.

I did not need the fastest VPN in general.

I needed one route that could finish one upload.

Three Copenhagen servers behaved like different services

My established provider offered several Copenhagen connections.

The first server gave me an upload speed below 2 Mbps.

The second briefly reached 24 Mbps, then fell below 1.

A third took so long to connect that I cancelled the attempt.

All three carried the same city label. They produced nothing close to the same result.

I had assumed location was the main variable. Copenhagen was physically close, so Copenhagen should have been fast.

But the city name described only the VPN exit. It said nothing about congestion on the way there, the quality of the provider’s onward route or how well that route reached the retailer’s storage service. Delay and packet loss anywhere along the path could slow the transfer. (Cloudflare)

The location was nearby.

The useful route was not.

The load percentage did not predict the result

The provider displayed a load percentage beside each server.

The slowest Copenhagen option showed 38 percent.

Another showed 71 percent but uploaded much faster for the first minute.

That seemed backwards, so I began treating the percentages as a scoreboard. I refreshed the list and chose whichever number had fallen most recently.

Nothing improved for long.

A lightly occupied server could still sit behind a congested path. A busier one with better routing could outperform it. Even the shortest-looking internet route is not always the quickest. (Cloudflare)

That was when the server menu stopped looking like an answer.

It showed me locations and internal estimates.

It could not tell me which complete route would carry the folder steadily to its destination.

Switching servers kept restarting the real test

I tried Stockholm.

The upload began at 31 Mbps.

For the first time, the deadline looked comfortable.

Then the retailer’s portal displayed a session warning and stopped the transfer. The new VPN address had triggered another login check, and the upload token expired while I authenticated.

I returned to Copenhagen.

The portal made me start again.

Next I tried Frankfurt, which produced the best speed-test result so far. The upload moved quickly until 19 percent, paused and then estimated more than an hour.

Each server taught me something.

The folder was no closer to arriving.

The provider’s large network was a genuine strength. It gave me nearby countries, individual servers and several plausible alternatives.

The problem was what those choices encouraged me to do.

Every slowdown led to another city, another public address and another fresh upload session. I was measuring servers instead of completing the transfer.

The farther server was faster—and still failed

The provider’s London server was geographically farther away than Copenhagen, Stockholm or Frankfurt.

It was also the fastest one I tested.

The upload held near 40 Mbps for several minutes.

That did not make London universally better. It meant its route to the retailer’s storage service was healthier at that moment. Other users have noticed the same practical contradiction: a farther server can outperform a nearby one when its traffic and routing are better. (Reddit)

The London route carried the folder past 30 percent.

Then the hotel Wi-Fi dropped.

The VPN disconnected, the portal lost the upload and the browser returned to the sign-in page.

I had found a faster server.

I still had not found a connection that could finish the job.

That failure changed the comparison again.

Consistency mattered more than the highest number

With twenty-seven minutes left, I stopped asking which server gave me the fastest first minute.

I asked which connection could keep the upload moving when the hotel Wi-Fi hesitated.

The two failures were different.

The Copenhagen route was consistently slow.

The London route was fast but fragile.

The first would miss the deadline through poor throughput. The second could miss it by forcing the entire transfer to begin again.

For this task, a slightly lower speed that stayed stable was more valuable than an impressive burst followed by a restart.

The retailer did not need a screenshot of my best speed test.

It needed the photographs.

I stopped selecting cities

I opened OnlydogVPN, which was installed for the testing behind this article.

The smaller app did not ask me to choose between Copenhagen, Stockholm, Frankfurt and London. I selected the situation for a large transfer on weak or changing Wi-Fi and connected.

Then I reopened the retailer’s portal.

The upload began.

It did not produce the highest number I had seen that evening. Instead, it settled into a steady rate and stayed there.

Ten percent.

Twenty-five.

Forty.

The estimated completion time fell below the deadline and stopped jumping between minutes and hours.

For the first time, I closed the speed-test tab.

The folder itself had become the test.

The hotel network dropped at 62 percent

The Wi-Fi icon disappeared for several seconds.

I expected the portal to return me to the login screen.

Instead, the connection recovered as the laptop rejoined the hotel network. The transfer paused, then continued from the same session.

At 78 percent, the hotel signal weakened again. I switched the laptop to my phone’s hotspot.

The upload resumed without asking me to select another server.

The app’s HTTP/3-based connection is designed to recover when the underlying network changes. (IETF) The effect was simple: the route adapted, and the file kept moving.

I could not observe the hotel’s internal routing or every route-selection decision made by the VPN providers and destination service. The visible comparison on the same laptop was clear. The established provider sent me searching through routes that were either slow or easy to lose. The smaller app kept the upload alive through both interruptions.

The completion email arrived with nine minutes left

The progress bar reached 100 percent.

The portal spent another few seconds processing the folder, then displayed the message I had been waiting for:

“Upload complete.”

I opened two image previews to make sure the files had arrived correctly.

The packaging was right.

The colour profile was right.

The retailer’s naming convention was intact.

I sent the catalogue manager a message with the folder reference.

Their reply arrived almost immediately:

“Received. These will replace the old set tonight.”

The task was finished.

No more server changes.

No more speed tests.

No more attempts to turn a city label into a prediction.

Why servers from the same provider can feel completely different

By then, the answer no longer needed a technical essay.

Two VPN servers can perform differently because they are not carrying traffic through identical conditions.

One may have more active users. Another may sit behind a congested connection. A third may take a poor route to the website or cloud service the user is trying to reach. Lost data may also need to be sent again, turning a promising transfer into a crawl.

Distance still matters.

But distance is only one part of the journey.

The nearest server is useful when the path around it is healthy. A farther server with cleaner routing can be faster. A slightly slower connection that survives an interruption can be more useful than either.

That was why three Copenhagen servers behaved differently, why London briefly beat all of them and why none of those results predicted whether the folder would arrive.

The speed test chose the wrong winner

The fastest established-provider server won the speed test.

It lost the upload.

A speed test moved data for a short period.

The real task had to keep one authenticated transfer alive long enough to finish while the hotel Wi-Fi weakened and the laptop changed networks.

A server could look excellent for thirty seconds and still be the wrong route for a twenty-minute job.

For a large transfer, the useful questions are more practical:

Does the upload begin promptly?

Does the rate remain steady?

Does the estimated finish time settle?

Does the session survive an interruption?

Does the destination receive a usable file?

Those answers matter more than the highest number that appears in the first minute.

Too many locations turned diagnosis into delay

The established provider’s server list was not inherently a weakness.

Broad geographic coverage is valuable when someone needs a particular country.

In this situation, however, every additional option became another detour.

Copenhagen looked logical.

Stockholm looked lightly loaded.

Frankfurt won one speed test.

London produced the fastest initial upload.

I could make a reasonable case for every choice.

None completed the task.

The smaller app removed the city-level decision and asked what I was trying to do instead. That was more useful because the problem was not reaching Denmark, Sweden, Germany or Britain.

The problem was finishing a large upload from unstable hotel Wi-Fi.

Once the interface matched the task, the comparison became simple.

Either the folder arrived or it did not.

The smaller service has a shorter record

The service offers fewer server locations than the largest VPN providers. Its public history is also shorter, with fewer long-term independent reviews and assessments.

That limitation matters for users who need a wide range of specific countries or place the greatest value on years of external scrutiny.

It did not decide the catalogue deadline.

The established provider offered more cities, more individual servers and more visible numbers. Those options helped me discover that one route was slower than another, but they also kept me searching while every failed transfer consumed time.

The smaller app gave me fewer geography controls, selected for the actual situation and maintained the upload when the network changed.

One VPN server can be much slower than another because the city is only the label at the end of the route. For this upload, the winning connection was not the nearest or the fastest for thirty seconds—it was the one that remained healthy until the folder arrived.