diff --git a/kafka/log_compacted_consumer.go b/kafka/log_compacted_consumer.go
index 4c1fa20c..9370e454 100644
--- a/kafka/log_compacted_consumer.go
+++ b/kafka/log_compacted_consumer.go
@@ -75,12 +75,13 @@ func (cfg *LogCompactedConfig) finalize() error {
 // record for each key, allowing for efficient storage and retrieval of the most
 // recent state of each key.
 type LogCompactedConsumer struct {
-	client  *kgo.Client
-	logger  *zap.Logger
-	process func(context.Context, *kgo.FetchesRecordIter) error
-	topic   string
-	ctx     context.Context
-	cancel  context.CancelFunc
+	client       *kgo.Client
+	logger       *zap.Logger
+	process      func(context.Context, *kgo.FetchesRecordIter) error
+	topic        string
+	ctx          context.Context
+	cancel       context.CancelFunc
+	fetchMaxWait time.Duration
 
 	mu      sync.RWMutex
 	started chan struct{}
@@ -117,13 +118,14 @@ func NewLogCompactedConsumer(cfg LogCompactedConfig,
 	}
 
 	lcc := LogCompactedConsumer{
-		topic:   cfg.Topic,
-		process: cfg.Processor,
-		logger:  cfg.Logger,
-		client:  client,
-		started: make(chan struct{}),
-		stopped: make(chan struct{}),
-		synced:  make(chan struct{}),
+		topic:        cfg.Topic,
+		fetchMaxWait: cfg.FetchMaxWait,
+		process:      cfg.Processor,
+		logger:       cfg.Logger,
+		client:       client,
+		started:      make(chan struct{}),
+		stopped:      make(chan struct{}),
+		synced:       make(chan struct{}),
 	}
 	lcc.ctx, lcc.cancel = context.WithCancel(context.Background())
 	return &lcc, nil
@@ -149,7 +151,14 @@ func (lcc *LogCompactedConsumer) Run(ctx context.Context) error {
 	go func() {
 		defer close(lcc.stopped)
 		close(lcc.started)
-		for {
+		// It's possible that the topic contains no records, in which case the
+		// consumer should be considered synced immediately. Since fetchMaxWait
+		// only works if the topic has records, we need to pass a context with
+		// a timeout to the first fetch.
+		ctx, cancel := context.WithTimeout(lcc.ctx, lcc.fetchMaxWait)
+		defer cancel()
+		lcc.consume(ctx)
+		for { // Business as usual: continue to fetch records until closed.
 			select {
 			case <-lcc.ctx.Done():
 				return // Exit the goroutine if the context is done.
@@ -167,7 +176,7 @@ func (lcc *LogCompactedConsumer) Run(ctx context.Context) error {
 }
 
 func (lcc *LogCompactedConsumer) consume(ctx context.Context) {
-	fetches := lcc.client.PollRecords(lcc.ctx, -1) // This means all buffered.
+	fetches := lcc.client.PollRecords(ctx, -1) // This means all buffered.
 	if fetches.IsClientClosed() {
 		lcc.logger.Info("kafka client closed, stopping fetch")
 		return
@@ -185,6 +194,15 @@ func (lcc *LogCompactedConsumer) consume(ctx context.Context) {
 		)
 	})
 	if fetches.Empty() {
+		// No records were returned across all partitions; consider the consumer
+		// synced because there is nothing to process to reach the current HWM.
+		select {
+		case <-lcc.synced:
+			// already synced
+		default:
+			lcc.syncDelta.Store(0)
+			close(lcc.synced)
+		}
 		return
 	}
 	lcc.mu.RLock()
@@ -260,6 +278,8 @@ func (lcc *LogCompactedConsumer) Close() error {
 
 // Healthy checks if the consumer is healthy by ensuring that it has had a full
 // sync and that the underlying Kafka client can ping a broker in the cluster.
+// If the first fetch returns no records, the consumer is considered synced
+// immediately. This may only happen when the topic has no records.
 //
 // This function can be used as a readiness probe.
 func (lcc *LogCompactedConsumer) Healthy(ctx context.Context) error {
diff --git a/kafka/log_compacted_consumer_test.go b/kafka/log_compacted_consumer_test.go
index 182f6ccd..83bce1c7 100644
--- a/kafka/log_compacted_consumer_test.go
+++ b/kafka/log_compacted_consumer_test.go
@@ -268,7 +268,7 @@ func TestLogCompactedConsumerSyncBehavior(t *testing.T) {
 			ClientID: "sync-test-consumer",
 		},
 		Topic:        topicName,
-		FetchMaxWait: 10 * time.Millisecond,
+		FetchMaxWait: time.Second,
 		Processor: func(_ context.Context, iter *kgo.FetchesRecordIter) error {
 			for !iter.Done() {
 				_ = iter.Next() // We just count records, don't need content
@@ -414,3 +414,43 @@ func TestLogCompactedConsumerStartStop(t *testing.T) {
 
 	assert.Error(t, consumer.Run(ctx), "Run should error after Close")
 }
+
+func TestLogCompactedConsumerHealthyOnEmptyFirstFetch(t *testing.T) {
+	topicName := "empty-first-fetch-topic"
+	cluster := newLogCompactedFakeCluster(t, topicName, 1)
+	var processedRecords int32
+	cfg := LogCompactedConfig{
+		CommonConfig: CommonConfig{
+			Brokers:  cluster.ListenAddrs(),
+			Logger:   zapTest(t),
+			ClientID: "empty-first-fetch-consumer",
+		},
+		Topic:        topicName,
+		FetchMaxWait: 100 * time.Millisecond,
+		MinFetchSize: 0,
+		Processor: func(_ context.Context, iter *kgo.FetchesRecordIter) error {
+			for !iter.Done() {
+				_ = iter.Next()
+				atomic.AddInt32(&processedRecords, 1)
+			}
+			return nil
+		},
+	}
+
+	consumer, err := NewLogCompactedConsumer(cfg)
+	require.NoError(t, err)
+	t.Cleanup(func() { consumer.Close() })
+
+	ctx, cancel := context.WithTimeout(context.Background(), time.Second)
+	defer cancel()
+	require.NoError(t, consumer.Run(ctx))
+
+	// With an empty topic, the first fetch should return no records and the
+	// consumer should be considered synced (healthy) without producing anything.
+	require.Eventually(t, func() bool {
+		return consumer.Healthy(ctx) == nil
+	}, time.Second, 50*time.Millisecond)
+
+	// Sanity check: no records were processed.
+	assert.Equal(t, int32(0), atomic.LoadInt32(&processedRecords))
+}