Segment events lost in serverless: await the flush before returning

If Segment events vanish from serverless functions, never trust fire-and-forget track calls: always await the flush before the handler returns. In AWS Lambda the same applies to code after callbackWaitsForEmptyEventLoop tricks; the queued flush is the thing that gets killed. Set explicit flushAt and flushInterval values so the only flush is the one you await, and log flush errors rather than swallowing them. In long-lived servers the background flush is fine, but any short-lived invocation must drain the queue synchronously.

Context: GitHub issues segmentio/analytics-node#303 and #309 (both closed): events tracked with analytics-node were silently never sent from Vercel serverless functions. The library queues events and flushes them asynchronously in the background; the serverless runtime finishes the invocation and freezes the process before the flush happens, so the queue dies with the function. The community-verified fix is to await the flush before returning the HTTP response: wrap track calls so their callbacks resolve promises, then await all of them (a ManagedSegmentClient finish pattern), or configure flushAt and flushInterval and explicitly flush at the end of the handler.

Matched source

Source: Published skill Original query: "Segment events lost in serverless: await the flush before returning" Key terms: await, before, events, flush, lost, returning, segment, serverless