diff --git a/src/Interpreters/GraceHashJoin.cpp b/src/Interpreters/GraceHashJoin.cpp
index 2a22833f29de..9d12525e408c 100644
--- a/src/Interpreters/GraceHashJoin.cpp
+++ b/src/Interpreters/GraceHashJoin.cpp
@@ -528,51 +528,68 @@ class GraceHashJoin::DelayedBlocks : public IBlocksStream
 
     Block nextImpl() override
     {
-        if (!tmp_result || tmp_result->isFinished())
         {
-            Block block;
-            size_t num_buckets = buckets.size();
-            size_t current_idx = buckets[current_bucket]->idx;
-
-            do
+            std::lock_guard lock(remaining_blocks_mutex);
+            if (!remaining_blocks.empty())
             {
-                    // One DelayedBlocks is shared among multiple DelayedJoinedBlocksWorkerTransform.
-                    // There is a lock inside left_reader.read() .
-                    block = left_reader.read();
-                    if (!block)
-                        return {};
-
-                    // block comes from left_reader, need to join with right table to get the result.
-                    Blocks blocks = JoinCommon::scatterBlockByHash(left_key_names, block, num_buckets);
-                    block = std::move(blocks[current_idx]);
-
-                    /*
-             * We need to filter out blocks that were written to the current bucket `B_{n}`
-             * but then virtually moved to another bucket `B_{n+i}` on rehash.
-             * Bucket `B_{n+i}` is waiting for the buckets with smaller index to be processed,
-             * and rows can be moved only forward (because we increase hash modulo twice on each rehash),
-             * so it is safe to add blocks.
-             */
-                    for (size_t bucket_idx = 0; bucket_idx < num_buckets; ++bucket_idx)
-                    {
-                        if (blocks[bucket_idx].rows() == 0)
-                            continue;
-
-                        if (bucket_idx == current_idx) // Rows that are still in our bucket
-                            continue;
-
-                        buckets[bucket_idx]->addLeftBlock(blocks[bucket_idx]);
-                    }
-            } while (block.rows() == 0);
-
-            ExtraBlockPtr not_processed;
-
-            tmp_result = hash_join->joinBlockWithStreamOutput(block, not_processed);
-
-            if (not_processed)
-                    throw Exception(ErrorCodes::LOGICAL_ERROR, "Unsupported hash join type");
+                    auto res = remaining_blocks.front();
+                    remaining_blocks.pop();
+                    return res;
+            }
+        }
+        Block block;
+        size_t num_buckets = buckets.size();
+        size_t current_idx = buckets[current_bucket]->idx;
+
+        do
+        {
+                // One DelayedBlocks is shared among multiple DelayedJoinedBlocksWorkerTransform.
+                // There is a lock inside left_reader.read() .
+                block = left_reader.read();
+                if (!block)
+                    return {};
+
+                // block comes from left_reader, need to join with right table to get the result.
+                Blocks blocks = JoinCommon::scatterBlockByHash(left_key_names, block, num_buckets);
+                block = std::move(blocks[current_idx]);
+
+                /*
+         * We need to filter out blocks that were written to the current bucket `B_{n}`
+         * but then virtually moved to another bucket `B_{n+i}` on rehash.
+         * Bucket `B_{n+i}` is waiting for the buckets with smaller index to be processed,
+         * and rows can be moved only forward (because we increase hash modulo twice on each rehash),
+         * so it is safe to add blocks.
+         */
+                for (size_t bucket_idx = 0; bucket_idx < num_buckets; ++bucket_idx)
+                {
+                    if (blocks[bucket_idx].rows() == 0)
+                        continue;
+
+                    if (bucket_idx == current_idx) // Rows that are still in our bucket
+                        continue;
+
+                    buckets[bucket_idx]->addLeftBlock(blocks[bucket_idx]);
+                }
+        } while (block.rows() == 0);
+
+        ExtraBlockPtr not_processed;
+        auto stream_output = hash_join->joinBlockWithStreamOutput(block, not_processed);
+
+        if (not_processed)
+                throw Exception(ErrorCodes::LOGICAL_ERROR, "Unsupported hash join type");
+        std::vector<Block> blocks;
+        while (!stream_output->isFinished())
+        {
+            blocks.push_back(stream_output->next());
+        }
+        std::lock_guard lock(remaining_blocks_mutex);
+        for (const auto & item : blocks)
+        {
+            remaining_blocks.push(item);
         }
-        return tmp_result->next();
+        auto res = remaining_blocks.front();
+        remaining_blocks.pop();
+        return res;
     }
 
     size_t current_bucket;
@@ -580,7 +597,8 @@ class GraceHashJoin::DelayedBlocks : public IBlocksStream
     InMemoryJoinPtr hash_join;
 
     AccumulatedBlockReader left_reader;
-    IBlocksStreamPtr tmp_result;
+    std::queue<Block> remaining_blocks;
+    std::mutex remaining_blocks_mutex;
 
     Names left_key_names;
     Names right_key_names;