I have a Worker running under `celld dev` that serves Cap'n Web RPC using capnweb's own `newWorkersRpcResponse` helper, no framework involved. Every RPC call made over the WebSocket transport hangs forever, it never resolves and never rejects. This is 100% reproducible, not intermittent. I tested five fresh connections in a row and all five hung on the very first call.

Plain HTTP batch RPC (`newHttpBatchRpcSession`) against the same worker works every time, so this is isolated to the WebSocket transport.

**Environment**

- celld 0.4.0, run via `celld dev` - capnweb ^0.12.0

**Repro**

```ts // worker.ts — run with `celld dev` import { RpcTarget, newWorkersRpcResponse } from 'capnweb';

interface PublicApi { hello(name: string): string; }

class Api extends RpcTarget implements PublicApi { hello(name: string) { return `Hello, ${name}!`; } }

export default { fetch(request: Request) { return newWorkersRpcResponse(request, new Api()); }, };

// client.ts — run against the worker above once `celld dev` is up import { newWebSocketRpcSession } from 'capnweb';

const stub = newWebSocketRpcSession('ws://127.0.0.1:9876/api'); console.log(await stub.hello('first call')); // never resolves ```

**Steps**

1. Start the worker above with `celld dev`. 2. Run the client against it. 3. `hello()` never resolves, no error, no rejection, on every attempt.

**Expected:** the call resolves, same as it does over HTTP batch RPC.

**Actual:** the WebSocket upgrade completes (101 response), but no RPC response ever comes back over it.

Also note: This is proven on the celld side, the same runs properly on `wrangler dev`. I also have a patch locally that works for `celld`, but PRs are disabled :(