Covers the module-level-caching anti-pattern for I/O objects (Request/Response/streams/database clients) in Workers. Does not cover the Cache API, which is the correct, supported way to cache actual HTTP responses across requests — this playbook is specifically about accidentally reusing the live object rather than its data.
Unverifiedno reproductions yetWhy this confidence?8 steps. Every step is written out below in full — the interactive version simply follows the branches for you.
Each Worker invocation has its own isolated execution context. Accessing a stream, Request, or Response body created in one invocation from a different invocation is disallowed and throws this error — it's the runtime protecting request isolation, not a transient failure.
What happens next
Search for a Request, Response, ReadableStream, or similar object assigned to a variable declared outside the fetch handler, then read again on a later invocation.
grep -n "^let \|^const \|^var " src/index.tsExpected result
A top-level variable such as `let cachedResponse = null;` that gets assigned inside fetch() and read on a subsequent call.What happens next
Extract and store the plain data (a string, parsed JSON, an ArrayBuffer) instead of the live Response/Request/stream. Construct a fresh Response from that data on each invocation. If the goal is genuine HTTP response caching across requests, use the Cache API instead, which is designed for exactly that.
let cachedData = null; // plain data, not a Response
export default {
async fetch(request, env, ctx) {
if (cachedData) return new Response(cachedData);
const res = await fetch("https://example.com/data");
cachedData = await res.text();
return new Response(cachedData);
},
};What happens next
The same error shows up when a database client — a Hyperdrive/pg connection, a driver instance — is created once at module scope and reused across invocations instead of being created fresh inside the handler.
grep -n "new Client\|new Pool\|connect(" src/**/*.tsExpected result
A client/connection constructed at module scope, outside fetch(), and reused across requests.What happens next
Fire two requests back to back and confirm neither throws, and that the second reflects fresh data rather than a stale cached I/O object.
curl -s https://your-worker.example.workers.dev/ && curl -s https://your-worker.example.workers.dev/Expected result
Both requests return normally with no "Cannot perform I/O" error.What happens next
Move client construction inside the request handler so each invocation gets its own. Hyperdrive's connection pooling already removes the connection-startup cost this pattern was usually trying to avoid, so there's no performance reason to hand-roll caching here.
export default {
async fetch(request, env) {
const client = new Client({ connectionString: env.HYPERDRIVE.connectionString });
await client.connect();
try {
return Response.json(await client.query("SELECT 1"));
} finally {
ctx.waitUntil(client.end());
}
},
};What happens next
If nothing at module scope holds an I/O object, look for the same pattern one level removed — an object stored in a class instance that itself persists across invocations (e.g. attached to a Durable Object instance field, which does persist between requests to that object).
I/O objects are now created fresh per invocation rather than shared across requests.
What would strengthen it: 6 more independent reproductions. Reproductions from 3 more distinct environments.