fix: eliminate race condition in TestBroadcastEvent_Delivery

[Copilot]

TestBroadcastEvent_Delivery was flaky under load: BroadcastEvent fired immediately after dialUnixWithRetry returned, before the server's serve goroutine had processed Accept()+trackConn(). The broadcast hit an empty conns map and was lost, causing a 5-second timeout.

Fix

• After dialing, poll srv.conns until the server has registered the connection before calling BroadcastEvent

for {
    srv.mu.Lock()
    tracked := len(srv.conns)
    srv.mu.Unlock()
    if tracked > 0 {
        break
    }
    select {
    case <-ctx.Done():
        t.Fatal("timed out waiting for server to track connection")
    case <-time.After(5 * time.Millisecond):
    }
}

Deterministic and no arbitrary sleeps — package ipc test access to unexported fields makes this straightforward.

---

📱 Kick off Copilot coding agent tasks wherever you are with GitHub Mobile, available on iOS and Android.

[Copilot]

Pull request overview

Stabilizes TestBroadcastEvent_Delivery in internal/ipc by ensuring the server has registered the newly-dialed connection before broadcasting, preventing occasional lost broadcasts due to a connection-tracking race.

Changes:

• Add a polling wait after dialing to ensure srv.conns contains the connection before calling BroadcastEvent.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.