fix: Add WaitGroup for process() goroutine completion

We need to ensure that the User.process() goroutine has finished
executing before we serialize the queue to disk. It is currently
possible that the goroutine is able to pop operations during the queue's
closure. This in turn can result in a race condition when accessing the
connection metadata.

internal/remote/user.go

@@ -38,6 +38,9 @@ type User struct {
 	closedLock sync.RWMutex
 
 	connMetadataStore connMetadataStore
+
+	// processWG is used to ensure we wait until the process goroutine has finished executing after we close the queue.
+	processWG sync.WaitGroup
 }
 
 // newUser constructs a new user with the given (IMAP) credentials.
@@ -61,6 +64,7 @@ func newUser(userID, path string, conn connector.Connector) (*User, error) {
 	// send connector updates along to the mailserver.
 	go user.forward(conn.GetUpdates())
 
+	user.processWG.Add(1)
 	// process remote operations on the operation queue.
 	go user.process()
 
@@ -79,6 +83,9 @@ func (user *User) Close() error {
 		return fmt.Errorf("failed to close queue: %w", err)
 	}
 
+	// Wait until any remaining operations popped by the process go routine finish executing
+	user.processWG.Wait()
+
 	if user.lastOp != nil {
 		ops = append([]operation{user.lastOp}, ops...)
 	}

internal/remote/user_queue.go

@@ -13,6 +13,8 @@ var ErrQueueClosed = errors.New("the queue is closed")
 // TODO: What should we do with operations that failed to execute due to auth reasons?
 // We might want to save them somewhere so we can try again after the user has logged back in.
 func (user *User) process() {
+	defer user.processWG.Done()
+
 	for {
 		op, ok := user.popOp()
 		if !ok {