The Speed Test Said 180 Mbps. The Page Still Wouldn’t Open.

The speed test needle jumped past 180 Mbps while the Reddit thread beside it remained a blank white page. I was on hotel Wi-Fi, trying to read an age-restricted discussion about a medication reaction before the nearby pharmacy closed.

The problem felt more urgent because stronger UK age checks had spread across adult and restricted services since July 2025. By 2026, most of the country’s largest pornography services had introduced age assurance or stopped serving UK visitors, while millions of checks were being completed across major platforms. (Ofcom)

That had pushed the VPN from a background privacy tool into the middle of a time-sensitive task. I was not measuring internet performance for curiosity. I needed one page to finish loading before deciding whether the symptoms on my arm could wait until morning.

The speed test appeared to prove that the connection had plenty of capacity.

What it did not prove was whether the connection could deliver a real webpage without repeatedly losing its place.

The test measured volume, not the waiting between requests

A speed test moves a large amount of data between the device and a nearby test server. Once that transfer begins, it can produce an impressive result even when ordinary browsing feels slow. (Fast)

A webpage asks more of the connection.

The browser first has to find the site, establish a secure connection and request the page. It may then contact several other services for photographs, scripts, fonts and embedded content. A delay in any one of those short exchanges can leave the reader staring at a white screen or a row of grey image boxes.

That was the mismatch in front of me.

The speed test had found one good route and kept data flowing through it. The Reddit page kept starting new requests, waiting for replies and retrying whatever the hotel network had dropped.

Bandwidth described how much the connection could carry once it was moving. It said little about how quickly each small part began or whether it arrived cleanly. Latency, packet loss and sudden changes in network quality can make a high-bandwidth connection feel unresponsive. (Cloudflare)

The 180 Mbps result was real.

It was simply answering the wrong question.

The established VPN looked fast after the delay

I was using a major provider with years of public history, a large support operation and servers across the world. It connected quickly, and its nearest route continued to produce speed-test results between 150 and 180 Mbps.

Then I opened a fresh browser window.

The first page waited nearly twenty seconds before anything appeared. Once it loaded, scrolling felt normal. The next unfamiliar site stalled again. Reddit eventually delivered the text, but its photographs remained grey.

I changed servers.

The next speed test was slightly slower but still more than sufficient for browsing. The initial page delay remained. A third server introduced a CAPTCHA before returning me to the same half-loaded discussion.

This pattern appears often in public support threads: a connection posts a respectable speed result, yet unfamiliar pages hesitate before loading and then behave normally once the first connection succeeds. (Reddit)

That short observation helped more than another speed test. The problem was not a lack of raw bandwidth.

The route was struggling with the repeated small exchanges that make a modern webpage usable.

The established provider was fast when the connection had settled. It was less convincing during the moments before the page settled at all.

Fixing the first delay exposed the second

I tried the quickest repairs.

First, I opened a private window so the browser would not reuse the earlier Reddit session. Then I changed DNS, hoping the delay occurred while the browser was trying to locate the site.

The title appeared faster.

The photographs still stalled.

That was useful progress because it separated two problems. The browser was finding the page more quickly, but the connection still struggled to retrieve everything the page needed afterward.

I could not observe the hotel network’s internal filtering, traffic queues or packet-handling rules. I could see the pattern: a large continuous download performed well, while short requests repeatedly hesitated or restarted.

The Wi-Fi signal also shifted whenever I moved away from the desk. The speed test eventually recovered from those brief changes. The browser often did not. It waited for missing images, retried connections and left parts of the page unfinished.

By then, changing servers had become another delay rather than a solution.

I had tried the nearest route, the fastest route and the app’s recommended route. All three produced good numbers. None completed the medical thread cleanly.

Peak speed was no longer the comparison standard.

The connection had to keep making progress when the hotel Wi-Fi became unreliable.

The smaller app finished the page before I checked its speed

OnlydogVPN had fewer locations, a shorter public history and fewer independent ratings than the established provider. Those were its clearest limitations.

Its interface started with the situation rather than a list of countries.

I selected the option intended for a weak or changing connection and reopened the Reddit thread in a clean browser session.

The title appeared.

Then the photographs.

Then the comments beneath them.

There was no long white pause before the first line, and the photographs did not remain trapped behind grey placeholders. I reached the section explaining which medication reactions were expected and which required professional advice.

Only after the page had loaded did I run another speed test.

The result was lower than the established provider’s best number.

It was also no longer important.

The thread had completed the task that the faster connection repeatedly failed to finish.

That changed the comparison immediately. The fastest route according to a test server had not been the fastest route to useful information.

The smaller app uses HTTP/3-based transport, additional obfuscation and recovery designed for unstable connections. The practical effect was simple: I did not have to choose a protocol or keep moving between servers. I selected the unstable-network option and returned to the page.

The technology stayed underneath the result, where it belonged.

The real test began when the Wi-Fi disappeared

The Reddit discussion linked to a medical-information page with a symptom form. I opened it and began entering the details.

Then the hotel Wi-Fi dropped.

My laptop switched to the phone hotspot I had used earlier that afternoon. With the previous setup, similar changes had left the VPN sitting on “reconnecting” while the browser stopped moving.

This time, the form paused briefly and continued.

The page did not restart. The answers I had already entered remained in place. I reached the recommendation to contact a pharmacist and copied the address before closing time.

That was the observable result I needed.

A speed test could show high throughput for thirty seconds on one stable connection. It could not prove that the route would survive a switch from weak hotel Wi-Fi to a hotspot halfway through the actual task.

The smaller app did.

Its recovery mattered more than another fifty megabits on the test screen.

The next page felt lighter for a second reason

After copying the pharmacy address, I opened its booking page. A blocked-request counter in the app began to rise.

Advertising and tracking services were attempting background connections around the page, and the app filtered a number of them before they completed.

That was not the main reason the Reddit thread loaded. The connection recovery had already solved the urgent problem.

The counter revealed a smaller advantage afterward. On an unstable network, every unnecessary request competes with the text, images and forms the reader actually needs. Reducing some of that background traffic gave the useful parts of the page less to compete with.

On a strong home connection, the difference might be difficult to notice. On hotel Wi-Fi that hesitated every few seconds, it made the browsing session feel cleaner.

It also gave me a reason to keep the app installed after the medical question was resolved.

The speed test had measured how quickly the connection could fill a pipe.

The completed pages showed how much useful work reached the other end.

Why pages can stall when the speed test looks fine

A strong download result means the connection can move data quickly under the conditions created by that particular test.

It does not guarantee that a new page will begin quickly, that every image server will respond cleanly or that the VPN will recover before the browser notices a network change.

That is why the symptoms can look contradictory:

The speed test starts slowly and then posts a large number. A familiar page works because its earlier connections are still available. A new page waits before showing anything. Text appears, but images from another service remain blank. The network changes, and the browser loses progress while the VPN reconnects.

The established provider delivered the larger speed-test result. It also left me waiting through repeated first-page delays and unfinished image loads.

The smaller app produced the result that mattered: the thread opened, the linked medical guidance loaded, and the form survived the move to another network.

For this problem, bandwidth was not the scarce resource.

The scarce resource was a connection that kept moving forward when the network stopped behaving perfectly.