Context
A reporting endpoint takes several seconds to finish and unrelated API calls become slower at the same time. Increasing the server size might change the symptom, but it does not explain whether request handling, a database query or report generation is responsible.
Measure a representative workload
Capture the request size, report dimensions, concurrency and cache state. Compare a small report with a large one while sending an ordinary lightweight request in parallel. Record endpoint latency and relevant process and dependency signals under the same conditions.
Node.js documents how lengthy work in event-loop callbacks or the worker pool can delay other clients. Merely declaring a function async does not establish that its computation is handled outside the JavaScript request path. Use measurements and profiling to identify the expensive operation.
Choose the execution boundary
A long export is often better treated as a job with a visible status than as a browser request that must remain connected. The design still needs explicit ownership: who can request it, which dataset it may read, how large it may become and when its output expires.
For CPU-heavy work, evaluate an appropriate bounded worker design or a separate processing service. For slow external dependencies, inspect timeouts, connection behavior and dependency capacity instead. Moving every delay to a new queue can conceal the original bottleneck and create another system to operate.
Protect the ordinary user journey
Set deliberate limits on export size and concurrent jobs. Define what happens when capacity is full: reject the request clearly, defer it or offer a smaller result. Avoid allowing one large customer task to consume every processing slot.
Retest with mixed traffic. A useful acceptance result is that the large export completes correctly while ordinary requests remain within the team's agreed performance objective. The objective should come from business requirements and measured capacity, not a universal response-time promise.
Leave an operating procedure
Document cancellation, retries, incomplete output cleanup and worker recovery. Store generated reports privately and check authorization when they are downloaded. Keep sensitive query values out of logs.
Clouvenix provides Node.js engineering and managed infrastructure support for application and server investigations.
Technical reference
Node.js guide to avoiding event-loop and worker-pool blocking.