Developer Tools

Why Your Apps Crash When Everyone Retries at Once

If you’ve ever cursed a website for crashing when you refreshed it, this explains why — and how to fix it.

Deep Dive

Ever refreshed a website only to see it crash entirely? That might not be a coincidence. A new study from researchers Rishabh Mehan and Jasmit Kaur Saluja found that when apps automatically retry failing requests — say, when a server is slow — it can backfire. Instead of fixing the problem, too many apps retrying at the same time overloads the system, making outages worse. They call this 'retry amplification,' where one small failure gets magnified into a full-blown crash.

The team studied 200 popular apps and found 41% used retry logic without proper safeguards. Worse, 60% of those apps didn’t even slow down their retries, meaning they kept hitting the server over and over as fast as possible. In simulations, this naive approach cut success rates from 55% to just 41%. Imagine if every time your Uber app failed to load, hundreds of users instantly refreshed it — the app would collapse.

The researchers propose a fix: 'Adaptive Retry Budgeting,' which limits how often apps can retry and spreads out the attempts. This keeps the system stable while still fixing temporary glitches. Their solution maintains success rates close to normal, even when parts of the system fail.

The takeaway? The next time your favorite app crashes, it might not be a hacker — it could be too many retries from everyone at once. Developers need to design retry policies as a system-wide rule, not just a quick patch per user.

Key Points
  • When apps retry failing requests at the same time, they can accidentally crash the whole service — even if the initial problem was small.
  • 41% of apps studied used retry logic without proper limits, making outages worse, not better.
  • Researchers propose 'Adaptive Retry Budgeting' — a smarter way to retry that prevents overload and keeps apps running.

Why It Matters

This explains why your apps sometimes crash when everyone refreshes at once — and how smarter retry rules could prevent it.

📬 Get the top 10 AI stories daily