diff --git a/cmd/axelard/cmd/vald/broadcaster/broadcast.go b/cmd/axelard/cmd/vald/broadcaster/broadcast.go
index ac80faecb..81b642650 100644
--- a/cmd/axelard/cmd/vald/broadcaster/broadcast.go
+++ b/cmd/axelard/cmd/vald/broadcaster/broadcast.go
@@ -62,9 +62,9 @@ func (b *Broadcaster) processBacklog() {
 		msgCount := 0
 		for {
 			// we cannot split a single task, so take at least one task and then fill up the batch
-			// until the size limit is reached or no new tasks are in the backlog
+			// until the size limit is reached
 			batchWouldBeTooLarge := len(batch) > 0 && msgCount+len(b.backlog.Peek().Msgs) > b.batchSizeLimit
-			if batchWouldBeTooLarge || b.backlog.Len() == 0 {
+			if batchWouldBeTooLarge {
 				break
 			}
 
@@ -72,6 +72,11 @@ func (b *Broadcaster) processBacklog() {
 
 			batch = append(batch, task)
 			msgCount += len(task.Msgs)
+
+			// if there are no new tasks in the backlog, stop filling up the batch
+			if b.backlog.Len() == 0 {
+				break
+			}
 		}
 
 		b.logger.Debug("high traffic; merging batches", "batch_size", msgCount)
@@ -369,10 +374,10 @@ func (bl *backlog) Peek() broadcastTask {
 }
 
 func (bl *backlog) Len() int {
-	bl.loadHead()
-
+	// do not block in this function because it might be used to inform other calls like Peek()
 	if len(bl.head.Msgs) == 0 {
-		return 0
+		// head is not currently loaded
+		return len(bl.tail)
 	}
 
 	return 1 + len(bl.tail)