refactor(server): clean up close listener and avoid writing 499 to dead socket

Wrap the long-poll handler in try/finally to detach the req.raw close
listener on every resolution path, and use reply.hijack() on the aborted
branch so Fastify does not attempt to write to a socket that's already
closed (which would otherwise emit per-disconnect log noise once the
watcher relaunch loop is busy). Behavior on the wire is unchanged for
the four documented status codes.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

node/src/server.ts

@@ -145,8 +145,9 @@ export async function buildServer(cfg: DaemonConfig, store: MailboxStore): Promi
       }
 
       const ac = new AbortController();
-      req.raw.on("close", () => ac.abort());
-
+      const onClose = (): void => ac.abort();
+      req.raw.once("close", onClose);
+      try {
         const result = await store.waitForMessage(name, timeoutS * 1000, ac.signal);
 
         if (result.kind === "message") {
@@ -162,9 +163,12 @@ export async function buildServer(cfg: DaemonConfig, store: MailboxStore): Promi
           reply.code(204);
           return reply.send();
         }
-      // aborted — client gone, no response needed (Fastify will swallow).
-      reply.code(499);
-      return reply.send();
+        // aborted — client gone; hand control to Fastify without writing to the dead socket.
+        reply.hijack();
+        return;
+      } finally {
+        req.raw.off("close", onClose);
+      }
     },
   );