Test your knowledge: fetchLater / Deferred Fetch

Question 1 of 6

`fetchLater()` differs from `fetch()` most fundamentally in that it…

The name is the contract: you describe the request now; the browser sends it later.

Answer options A returns no promise for the response — delivery is deferred to a browser-chosen moment B cannot send request bodies C only works inside Service Workers D is synchronous like sendBeacon

fetchLater() queues a Deferred Fetch and returns a record (with an abort handle and an activate() method) — never a response promise. The browser chooses the transmission moment under its own scheduling, with a much larger budget than beacons.

Review this lesson

Question 2 of 6

The Deferred Fetch quota is…

A tab-wide pool of its own — and NOT shared with the keepalive pool.

Answer options A the same 64 KiB pool, shared with beacons B 640 KiB — a separate pool with its own accounting C unlimited for activateAfter-scheduled requests D 256 KiB aggregate, reviving the old myth

fetchLater draws on its own deferred-fetch quota — 640 KiB per tab in the spec's (non-normative) quota note, with at most 64 KiB concurrently per reporting origin — separate from the 64 KiB keepalive pool. A QuotaExceededError there does not collide with in-flight beacons, which is exactly why the fallback ladder works.

Review this lesson

Question 3 of 6

When does a queued deferred fetch fire?

The spec's wait algorithm has exactly one unconditional trigger — and it is not page-hidden.

Answer options A At page-hidden — the same guarantee sendBeacon carries B At fetch-group termination; with activateAfter set, also after the delay or on background/long-hidden C When the page calls a dispatch() method on the result D When the server polls a registration endpoint

Without activateAfter, deferred records are processed when the fetch group is terminated — the page goes away. With activateAfter set, the wait also ends when the delay elapses or the browser is about to lose the chance to execute scripts (backgrounding, long-hidden documents). There is no dispatch() method, and no unconditional hidden trigger — that guarantee belongs to sendBeacon alone.

Review this lesson

Question 4 of 6

`activateAfter` lets a page…

It is a scheduling floor for the deferred transmission — useful for batch pings.

Answer options A cancel a deferred fetch after N milliseconds B schedule the deferred send no earlier than N milliseconds in the future C retry a failed deferred fetch after N milliseconds D keep the response body alive for N milliseconds

activateAfter: 30000 means the deferred fetch may not fire in the first 30 s — letting the page batch later events into one payload before the scheduler transmits. It composes with the deferred-fetch budget for exactly-once batch telemetry.

Review this lesson

Question 5 of 6

fetchLater replaces sendBeacon when…

Newest feature, biggest quota — and the smallest install base.

Answer options A the browser ships it AND your payloads are large or batched — but feature detection still gates the choice in 2026 B the collector is same-origin only C the page never navigates away D cookies must be omitted

fetchLater's browser-owned scheduling and its own deferred-fetch budget suit batch telemetry — but it is the newest API in the family and lacks sendBeacon's flush-on-hidden guarantee, so the recipe's ladder checks fetchLater → sendBeacon → keepalive fetch, and only promotes it when present.

Review this lesson

Question 6 of 6

Why does the fetchLater → sendBeacon → fetch(keepalive) fallback ladder work, quota-wise?

The lesson's warning: a false from rung 2 will repeat at rung 3 — same pool.

Answer options A All three share one pool, so failure order is deterministic B The 640 KiB pool is separate from the 64 KiB keepalive pool — but rungs 2 and 3 DO share the 64 KiB C Each rung has a strictly larger quota than the last D Quotas do not apply to fallback paths

fetchLater's QuotaExceededError is about the deferred-fetch quota — falling back to sendBeacon is meaningful. But sendBeacon's false and fetch(keepalive)'s rejection both reflect the SAME 64 KiB keepalive pool: the last two rungs fail together for quota reasons, and the ladder only helps for coverage- or feature-specific failures.

Review this lesson

Test your knowledge: fetchLater / Deferred Fetch