Request Coalescer

Problem

Build a proxy endpoint that coalesces concurrent identical requests to a slow backend, ensuring only one backend call is made even when many clients request the same resource simultaneously.

Background

When a cache entry expires and many clients request the same resource at once, a naive implementation sends N requests to the backend — a thundering herd or cache stampede. Request coalescing (also called request deduplication or single-flighting) ensures only one request goes to the backend; all other concurrent callers wait for and share the result.

This pattern is critical in CDNs, API gateways, and any service that proxies to a slower backend.

Requirements

• Implement GET /api/data/:key — a proxy endpoint that fetches from a simulated slow backend
• The backend is available at http://localhost:3001/backend/:key (provided by the harness)
• When multiple concurrent requests arrive for the same key, only one request should be made to the backend
• All concurrent callers for the same key should receive the same response
• Different keys should be fetched independently
• After a request completes, subsequent requests for the same key should call the backend again (no persistent caching — just coalescing in-flight requests)

Your Server

• Start an HTTP server on the port specified by the PORT environment variable (default: 3000)
• The harness starts a simulated slow backend on port 3001 before your server starts
• The backend responds with { "key": "", "value": "", "timestamp": } after a 500ms delay

API Contract

`GET /health`

Health check endpoint. Return 200 when your server is ready.

`GET /api/data/:key`

Fetch data for the given key from the backend, with request coalescing.

Response:

• Status: 200
• Body: The backend's response { "key": "", "value": "", "timestamp": }