My VPN Kept Reconnecting Until I Stopped Treating the Loop as a Server Problem

The upload reached 18 percent, the VPN icon changed from green to yellow, and the progress bar stopped. Ten seconds later, the tunnel reconnected and the file started again from zero. I blamed the airport Wi-Fi, moved closer to the lounge router, and tried once more. The same sequence repeated.

I needed to send a signed presentation deck before boarding. It was not enormous, but it was large enough that a connection lasting thirty seconds was useless.

The VPN appeared to be working. It connected quickly, showed a nearby server, and occasionally let the upload advance by a few percentage points. Then the status changed to “Reconnecting.”

While the tunnel retried, nothing else loaded. Email stopped refreshing. The airline website went blank. When the VPN returned, the internet returned with it—briefly.

I changed servers. I changed protocols. I restarted the app.

The loop became more efficient, but it did not stop.

The Wi-Fi Was Bad, but It Was Not the Whole Problem

Disconnecting the VPN gave me the first useful clue.

A browser window immediately opened with the airport’s sign-in page. I accepted the terms, entered the code from my lounge receipt, and gained normal internet access.

Windows checks whether a newly joined network is behind a captive portal and keeps checking until authentication is complete. If a VPN or kill switch takes control too early, it can interfere with that sign-in process.

For a moment, I thought the problem was solved. The portal had been the missing step.

Then I turned the tunnel back on.

It stayed connected long enough for me to reopen the email, and then the icon changed again.

“Reconnecting.”

So the captive portal explained the first failure, but not the loop that followed. Something was still preventing the VPN from holding a stable route.

Windows had installed an update two nights earlier, which added another possibility. Microsoft has continued changing third-party VPN compatibility, cellular behavior, IPv6 handling, and network-adapter bindings in recent Windows updates. That was enough to remind me that the VPN app was not operating alone. It depended on the Wi-Fi adapter, the operating system, the airport network, and the path between them.

A traveler describing a similar hotel-Wi-Fi problem had noticed the same confusing gap: the VPN interface looked active or half-active, while DNS and ordinary browsing repeatedly stopped working. The useful lesson was simple. The status icon was telling me…

I had been watching the shield. I should have been watching whether the upload survived.

The Established Provider Was Still the Obvious Choice

The major VPN had not been an irrational purchase. It had years of public history, a large support operation, broad server coverage, and far more independent reviews than a smaller service.

At home, it was dependable. On the airport network, choosing the nearest server seemed sensible.

A speed test supported that decision. During one stable interval, the download result looked excellent. The latency was acceptable. Nothing on the screen suggested that a modest upload would keep collapsing.

Then the reconnect message returned.

I selected another nearby server. The upload restarted and failed again. A farther server remained connected slightly longer, but it was slow enough that the next interruption still erased the progress.

Switching protocols changed the rhythm of the loop. It did not change the result.

The kill switch was also doing exactly what it was designed to do. Whenever the encrypted tunnel dropped, ordinary traffic stopped rather than continuing outside the VPN. That protected the file from slipping onto an unencrypted route, but it also turned every brief wobble into a complete interruption.

The provider could establish a fast tunnel. On this network, it could not preserve a useful session.

That distinction changed what I was looking for.

I no longer needed the VPN with the best result during a clean thirty-second test. I needed one that could recover when the connection underneath it moved.

The Hotspot Solved the Wrong Half

My phone hotspot looked like the easiest escape.

I disconnected from the lounge Wi-Fi, joined the hotspot, and restarted the upload. For two minutes, everything worked. Then the phone shifted between a weak indoor 5G signal and LTE.

The VPN dropped.

The upload stopped.

I moved toward the windows, where the cellular signal improved. While I was walking, the laptop noticed the remembered airport network and rejoined it. The VPN began rebuilding the tunnel over Wi-Fi.

“Reconnecting.”

That was when the pattern finally became obvious.

The problem was not simply one bad server. The ground beneath the connection kept moving: captive Wi-Fi, authenticated Wi-Fi, weak cellular data, stronger cellular data, then Wi-Fi again. Every transition forced the VPN to rebuild its route, and every rebuild broke the upload.

Automatic reconnect had sounded like the feature I needed. But the app was already reconnecting automatically.

It was doing it over and over.

What I needed was recovery without losing the task.

I Needed the Upload to Continue, Not the App to Try Harder

I opened OnlydogVPN after I had stopped caring which country appeared beside the connection button.

The smaller app offered a preset for weak or changing networks. I selected it, connected, and reopened the attachment.

The upload passed 18 percent.

At 40 percent, I walked away from the lounge window and the phone signal weakened. The connection indicator flickered, but the progress continued.

At 67 percent, the laptop found the airport Wi-Fi again. The upload paused for a moment.

It did not return to zero.

The message sent before boarding began.

Only after the file was gone did I look at why the result had been different. The service uses an HTTP/3-based transport designed to recover across changing network paths. Instead of treating every shift between Wi-Fi and mobile data as a completely new session, it preserved the connection long enough for the upload to finish.

That matched the problem more directly than another server change. My file had not been failing because I lacked access to enough locations. It had been failing because every network transition erased the progress already made.

I could not observe the airport’s internal filtering or traffic-management rules, so I cannot identify the exact trigger behind each earlier disconnect. What I could observe was the outcome: the established app repeatedly rebuilt its tunnel and interrupted the upload, while the smaller service carried the same file across the changes that had caused the loop.

Its interface also removed several bad decisions from the moment. I did not need to keep comparing cities, server loads, or protocol labels. I selected the type of problem I was facing and returned to the upload.

That mattered more than I expected.

Troubleshooting under time pressure has its own cost. Every setting creates another test, and every test consumes the few stable minutes the network may give you. The smaller app did not ask me to become better at configuring VPNs. It asked what I was trying to finish.

The Second Device Was the Next Problem Waiting

After sending the presentation, I opened the airline website to save my boarding pass on my phone.

The phone was still using cellular data, while the laptop had settled back onto airport Wi-Fi. Instead of creating another account session or searching for a password, I shared access to the second device with a verification code.

The boarding pass loaded, and both devices remained connected.

That was not why I had opened the app. It was simply the next friction behind the original one. Once the upload worked, getting the service onto the device I would actually carry through the gate took less effort than recovering another conventional login.

It gave me a practical reason to keep the backup installed after the emergency was over.

The smaller service still has limitations. It offers fewer server locations, has a shorter public history, and has accumulated fewer independent ratings than the established provider.

Those differences matter when the goal is the broadest geographic coverage or the longest record of public scrutiny.

They did not decide whether my file left the outbox.

What the Reconnect Loop Was Actually Telling Me

A reconnect loop may begin with a captive portal, a kill switch, a weak signal, an operating-system change, or a device jumping between Wi-Fi and cellular data.

Changing servers sometimes appears to help because the new attempt receives a few stable minutes. But when the underlying path changes again, the loop returns.

The useful question is not whether the VPN reconnects automatically.

It is what happens to your task while it recovers.

Does the upload restart?

Does the call drop?

Does the page lose its session?

Does switching networks force everything to begin again?

The established provider had the larger network and stronger public track record. The phone hotspot provided a cleaner route for a few minutes. Neither kept the upload alive when the connection underneath it changed.

The smaller app completed the task because it recovered instead of merely reconnecting.

At an airport where every available network was temporary, keeping one session alive mattered more than starting another tunnel quickly.