10 Popular Front-End Interview Answers Ran Against Real Code: 6 Are Flat Wrong
Candidates who recite these answers will fold on the second follow-up question, and production code built on the wrong units or limits will misbehave. The gap between memorized numbers and measured behavior is exactly what interviewers probe with boundary questions.
The most circulated front-end interview answers fall apart when executed. localStorage's quoted 5MB limit is actually 5,222,400 characters (UTF-16 code units), so ASCII content stores about 10MB. The 20-cookie-per-domain figure is IE6's limit; Chromium hard-codes 180 and silently evicts the oldest. GET URL length is not constrained by HTTP at all — Chrome sent a 1MB URL and got a 200 response; 2048 is an IE relic.
Promise.all does not cancel sibling requests when one rejects; the others finish their HTTP calls. Object.assign is always a shallow copy, and JSON-based deep copy drops undefined, turns Date into a string, and throws on circular references. The claim that "let is not hoisted" is contradicted by the error "Cannot access 'x' before initialization" — the binding exists, it's just in the temporal dead zone. Even the data-type count is stale: six primitives predates bigint, so the correct total is seven primitives plus object.
The failures cluster into three causes: IE-era facts copied without their qualifiers, answers written without ever running the code, and spec phrases like "5 MiB per origin" degraded into "5MB" in retransmission. All tests ran on Node 24.14.1 and local Chrome, with sources archived and scripts reproducible.
The three failure modes — fossilized IE facts, code never executed, and qualifiers dropped in retransmission — explain why question banks drift so far from reality.
An interviewer's second-round question probes the boundary of an answer; reciting a conclusion without its conditions is what separates candidates who ran code from those who copied text.
Spec language like "5 MiB per origin" and "IE6 or lower" contains the exact qualifiers that later retellings strip away, making the original document the only reliable check.