Users report intermittent UI glitches in different browsers—how would you troubleshoot

Reproduce systematically — gather environment data, use RUM/session replay, test across browsers. Intermittent + browser-specific points to race conditions, CSS/JS engine differences, third-party scripts, or extensions. Isolate, instrument, and narrow with a binary search.

7 min read·~12 min to think through

"Intermittent" + "browser-specific" is a hard combination — you can't fix what you can't see. The whole game is turning it reproducible.

1. Gather data — make the invisible visible

You can't reproduce it locally, so collect from the field:

Exact environment from reporters — browser + version, OS, device, screen size, network, extensions, steps.
RUM + error monitoring (Sentry, etc.) — JS errors segmented by browser/version. Is one engine over-represented?
Session replay (LogRocket, FullStory) — watch the glitch happen in the user's actual session. Often the single most useful tool here.
Look for patterns — specific browser? viewport? slow network? logged-in vs out? after a specific action?

2. Reason about likely causes (intermittent + cross-browser)

Race conditions — async ordering, effects firing in unexpected order, requests resolving out of order, animations vs data. "Intermittent" screams timing.
Browser engine differences — CSS support/quirks (fl/grid edge cases, :has(), container queries), JS API availability, default styles, event timing, font rendering.
Layout timing — measuring layout before it settles; CLS; race between fonts/images loading and JS.
Third-party scripts / extensions — ad blockers, password managers, translation tools mutating the DOM; analytics scripts loading non-deterministically.
Caching/versioning — stale chunks after a deploy, stale service worker.
State leakage — uncleaned listeners, stale closures, memory growth over a long session.

3. Reproduce and isolate

Match the environment — exact browser version, real device (BrowserStack/Sauce Labs for the ones you don't have), throttled network/CPU.
Test with extensions disabled vs enabled; incognito.
Binary search the cause — disable third-party scripts, toggle features/flags, bisect recent deploys (git bisect against the regression window).
Add targeted instrumentation — log timing/ordering around the suspected area, ship it, watch RUM. Make the intermittent visible.
Stress it — rapid interactions, slow network, repeated actions to force the race.

4. Fix and verify

Fix the root cause — guard the race (cancel stale requests, sequence effects, await properly), add the CSS fallback/prefix, defensively handle the third-party DOM mutation.
Cross-browser test the fix; add a regression test (E2E across browsers via Playwright).
Confirm via RUM that the error rate drops; follow up with reporters.

The framing

"Intermittent and browser-specific tells me race condition or browser-engine difference. I can't fix what I can't see, so step one is RUM + session replay + environment data to make it reproducible. Then match the environment exactly, binary-search the cause (extensions, third-party scripts, recent deploys), instrument the suspect area, fix the root cause, and add a cross-browser regression test."

Follow-up questions

•Why does 'intermittent' often point to a race condition?
•How does session replay help with bugs you can't reproduce?
•How would you isolate whether a browser extension is the cause?
•How do you binary-search for which deploy introduced a regression?

Common mistakes

•Trying to fix it without reproducing it first.
•Only testing in your own browser.
•Ignoring extensions and third-party scripts as suspects.
•Patching the symptom instead of the timing/root cause.
•No regression test, so it comes back.

Performance considerations

•Many intermittent glitches are timing/perf-related — layout measured too early, races between fonts/images/JS, jank under throttled CPU. Reproducing under throttling often surfaces them.

Edge cases

•Bug only on a specific browser version that later auto-updates away.
•Caused by a user's extension you can't control.
•Stale cached chunks after a deploy.
•Only reproduces on slow networks or low-end CPUs.

Real-world examples

•A glitch traced via session replay to a password-manager extension injecting DOM.
•An intermittent layout flash fixed by awaiting font load before measuring; git bisect pinpointing the deploy.

Senior engineer discussion

Seniors immediately read the clues — intermittent => race/timing, browser-specific => engine difference or extension/third-party — and prioritize making it reproducible via RUM, session replay, and environment data. They isolate by binary search (extensions, scripts, deploy bisect), instrument the suspect, fix the root cause, and lock it with a cross-browser regression test.