diff --git a/modin/experimental/engines/omnisci_on_ray/frame/data.py b/modin/experimental/engines/omnisci_on_ray/frame/data.py
index e1329d9f377..dc526ac3efc 100644
--- a/modin/experimental/engines/omnisci_on_ray/frame/data.py
+++ b/modin/experimental/engines/omnisci_on_ray/frame/data.py
@@ -61,6 +61,7 @@ def __init__(
         dtypes=None,
         op=None,
         index_cols=None,
+        uses_rowid=False,
     ):
         assert dtypes is not None
 
@@ -93,6 +94,8 @@ def __init__(
         if partitions is not None:
             self._filter_empties()
 
+        self._uses_rowid = uses_rowid
+
     def id_str(self):
         return f"frame${self.id}"
 
@@ -412,6 +415,7 @@ def _union_all(
         for frame in [self] + other_modin_frames:
             aligned_index = None
             exprs = OrderedDict()
+            uses_rowid = False
 
             if not ignore_index:
                 if frame._index_cols:
@@ -425,6 +429,7 @@ def _union_all(
                     aligned_index = ["__index__"]
                     exprs["__index__"] = frame.ref("__rowid__")
                     aligned_index_dtypes = [_get_dtype(int)]
+                    uses_rowid = True
                 aligned_dtypes = aligned_index_dtypes + new_dtypes
             else:
                 aligned_dtypes = new_dtypes
@@ -442,6 +447,7 @@ def _union_all(
                     dtypes=aligned_dtypes,
                     op=aligned_frame_op,
                     index_cols=aligned_index,
+                    uses_rowid=uses_rowid,
                 )
             )
 
@@ -628,7 +634,7 @@ def _execute(self):
         if isinstance(self._op, FrameNode):
             return
 
-        # MaskNode requires rowid which is available for executed frames only.
+        # Some frames require rowid which is available for executed frames only.
         # Also there is a common pattern when MaskNode is executed to print
         # frame. If we run the whole tree then any following frame usage will
         # require re-compute. So we just execute MaskNode's operands.
@@ -640,12 +646,18 @@ def _execute(self):
         self._partitions = new_partitions
         self._op = FrameNode(self)
 
+    def _require_executed_base(self):
+        if isinstance(self._op, MaskNode):
+            return True
+        return self._uses_rowid
+
     def _run_sub_queries(self):
         if isinstance(self._op, FrameNode):
             return
 
-        if isinstance(self._op, MaskNode):
-            self._op.input[0]._execute()
+        if self._require_executed_base():
+            for op in self._op.input:
+                op._execute()
         else:
             for frame in self._op.input:
                 frame._run_sub_queries()
diff --git a/modin/experimental/engines/omnisci_on_ray/test/test_dataframe.py b/modin/experimental/engines/omnisci_on_ray/test/test_dataframe.py
index e242f8b5b53..6d1be2b9d4a 100644
--- a/modin/experimental/engines/omnisci_on_ray/test/test_dataframe.py
+++ b/modin/experimental/engines/omnisci_on_ray/test/test_dataframe.py
@@ -149,6 +149,20 @@ def test_concat_many(self):
 
         df_equals(ref, exp)
 
+    def test_concat_agg(self):
+        def concat(lib, df1, df2):
+            df1 = df1.groupby("a", as_index=False).agg(
+                {"b": "sum", "d": "sum", "e": "sum"}
+            )
+            df2 = df2.groupby("a", as_index=False).agg(
+                {"c": "sum", "b": "sum", "f": "sum"}
+            )
+            return lib.concat([df1, df2])
+
+        run_and_compare(
+            concat, data=self.data, data2=self.data2,
+        )
+
 
 class TestGroupby:
     data = {