diff --git a/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/WALEntryStream.java b/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/WALEntryStream.java
index 8eec02a661e8..14e95852bdb9 100644
--- a/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/WALEntryStream.java
+++ b/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/WALEntryStream.java
@@ -343,11 +343,13 @@ private HasNext tryAdvanceEntry() {
       case EOF_WITH_TRAILER:
         // in readNextEntryAndRecordReaderPosition, we will acquire rollWriteLock, and we can only
         // schedule a close writer task, in which we will write trailer, under the rollWriteLock, so
-        // if beingWritten == true, it means that the reader is opened before writing the trailer,
-        // so we can not get a EOF_WITH_TRAILER return value while reading from this reader.
-        // this is very important for make sure that we will not consume the logQueue to empty for a
-        // normal replication queue
-        assert !beingWritten;
+        // typically if beingWritten == true, we should not reach here, as we need to reopen the
+        // reader after writing the trailer. The only possible way to reach here while beingWritten
+        // == true is due to the inflightWALClosures logic in AbstractFSWAL, as if the writer is
+        // still in this map, we will consider it as beingWritten, but actually, here we could make
+        // sure that the new WAL file has already been enqueued into the logQueue, so here dequeuing
+        // the current log file is safe.
+        assert beingWritten && logQueue.getQueue(walGroupId).size() > 1;
         // we have reached the trailer, which means this WAL file has been closed cleanly and we
         // have finished reading it successfully, just move to the next WAL file and let the upper
         // layer start reading the next WAL file