WordPress
Site Slow Under Load? A Real Case Study in Why
The Short Version
- A client’s site kept slowing to a crawl, sometimes timing out, but only under load: content uploads, busy stretches. It looked fine the rest of the time.
- The instinct is “upgrade the server.” Here the server had plenty of horsepower to spare. A bigger server would have fixed nothing.
- The real cause was one setting: the server was cold-starting a fresh worker process for every burst instead of keeping any ready.
- We reconfigured it to keep a small pool of workers warm at all times. No added hosting cost.
- The same page went from over three seconds to about a tenth of a second, and bursts that used to time out now finish in under half a second.
What happened
A client’s site had a problem that’s uniquely frustrating to chase down: it worked fine, until it didn’t. It would slow to a crawl, and sometimes time out completely, but only under load. Staff uploading content. Busy stretches. Then it would go right back to normal, like nothing happened.
Intermittent problems are the worst kind. If a site is broken all the time, you fix it and you’re done. If it’s broken sometimes, you’re stuck guessing whether whatever you just changed actually worked, or whether it just happened not to fail during your test.
The hosting looked fine on paper. The site looked fine most of the time. But every so often, it stalled.
Why a site slow under load doesn’t always mean bad hosting
When a site slows down under load, there are two instincts almost everyone reaches for first: the server is too small, or a plugin is misbehaving. Sometimes that’s exactly right.
Here, it wasn’t. The server had memory and processing power to spare, sitting nearly idle in between the stalls. Throwing a bigger, pricier server at this problem would have changed nothing, because the server was never actually the bottleneck.
Worth knowing
A server that looks idle most of the time and only chokes in short bursts is usually telling you the problem is a setting, not a shortage. More horsepower does not fix a configuration issue.
What we found: why this site was slow under load
The culprit was a single setting in how the server ran WordPress behind the scenes.
It was configured to spin up a fresh worker process for each burst of activity, on demand, then shut it back down once things went quiet. On paper that sounds efficient. Why keep something running if nothing needs it?
In practice, a sudden burst, like a content upload firing off dozens of background requests at once, forced the server to cold-start those workers from scratch, every time. Cold-starting is slow. While it happened, requests piled up behind it and started timing out.
The proof was clean. The exact same page loaded in a fraction of a second when a warm worker was already available, and took several seconds when the server had to cold-start one first. Same page, same content, same server. The only difference was whether a worker was already warm.
What we did about it
- Confirmed the real cause with a before-and-after test. We measured the same page under a warm worker and under a cold start, side by side, so the fix was based on proof, not a guess.
- Reconfigured the server to keep a small pool of workers warm at all times. Requests now get an available worker instead of waiting on one to spin up from nothing. No added hosting cost, no bigger server, no plan upgrade.
The results, verified
The stalls disappeared. We didn’t just take that on faith, we measured it.
30x
Faster page load
<0.5s
Burst response time
$0
Added hosting cost
The same page that once took over three seconds now loads in about a tenth of a second. A burst of simultaneous requests, the exact scenario that used to time out, now all complete in under half a second.
Curious where your own site stands under load? Our free Site Intel audit scores your site across 11 sections and flags issues like this one before they turn into a support ticket.
What a slow WordPress site actually needs
“The site is slow, upgrade the server” is one of the most common, and most expensive, pieces of advice in WordPress. Sometimes it’s right. Often the real problem is a configuration detail that no amount of extra hardware will ever touch.
Knowing where to look is the difference between a five-minute change and a bigger hosting bill that doesn’t solve anything.
That’s the job. Not guessing. Looking.
Frequently asked questions
Why does my WordPress site only slow down under load?
Intermittent slowdowns usually point to a bottleneck that only shows up when several things happen at once, like a burst of requests during a content upload or a traffic spike. If the site runs fine at rest, the server size is rarely the actual problem.
Is upgrading my hosting plan the right fix for a slow WordPress site?
Sometimes, but not always. If your server is sitting mostly idle between slowdowns, a bigger plan won’t help, because you already have more capacity than you’re using. The issue is usually how that capacity gets used, not how much of it you have.
What is a “cold start” and why does it slow down a website?
A cold start happens when a server has to spin up a new worker process from scratch to handle a request instead of using one that’s already running. Spinning up a fresh process takes real time, and while it happens, requests queue up behind it. A warm, already running worker skips that delay entirely.
How can I tell if my site’s slowdown is a configuration issue rather than a hosting size issue?
Check whether the server has room to spare when the site is running normally. If resource usage is low most of the time and only spikes during the slow periods, that points to a configuration detail rather than a genuine capacity shortage. A before-and-after load test on the same page is the clearest way to confirm it.
Will keeping server workers warm increase my hosting costs?
Not necessarily. Keeping a small pool of workers warm instead of letting them spin down completely is a configuration change, not a resource upgrade. In this case it required no added hosting cost at all.
Is Your Site Slow For a Reason Nobody’s Found Yet?
Intermittent slowdowns are easy to misdiagnose and expensive to guess your way through. Our care plans include performance monitoring and real troubleshooting from people who look for the actual cause before recommending a fix. We manage 105+ active client sites, and this is exactly the kind of thing we catch.
