diff --git a/tests/tape_tests/data/object/test_object2.parquet b/tests/tape_tests/data/object/test_object2.parquet
new file mode 100644
index 00000000..9f5fac64
Binary files /dev/null and b/tests/tape_tests/data/object/test_object2.parquet differ
diff --git a/tests/tape_tests/data/source/test_source2.parquet b/tests/tape_tests/data/source/test_source2.parquet
new file mode 100644
index 00000000..e6697d4f
Binary files /dev/null and b/tests/tape_tests/data/source/test_source2.parquet differ
diff --git a/tests/tape_tests/test_ensemble.py b/tests/tape_tests/test_ensemble.py
index 147d14b5..a389c10d 100644
--- a/tests/tape_tests/test_ensemble.py
+++ b/tests/tape_tests/test_ensemble.py
@@ -2054,11 +2054,24 @@ def my_mean(flux, time, dup_time, band):
 
 @pytest.mark.parametrize("on", [None, ["ps1_objid", "filterName"], ["filterName", "ps1_objid"]])
 @pytest.mark.parametrize("func_label", ["mean", "bounds"])
-def test_batch_by_band(parquet_ensemble, func_label, on):
+def test_batch_by_band(func_label, on):
     """
     Test that ensemble.batch(by_band=True) works as intended.
     """
 
+    # TODO(wbeebe): This is a temporary workaround for https://github.com/lincc-frameworks/tape/issues/434 to avoid contention when reading data
+    # from the test files in parallel. This is the same as `parquet_ensemble` but uses copies of the parquet files referenced there.
+    parquet_ensemble = Ensemble(client=False)
+    parquet_ensemble.from_parquet(
+        "tests/tape_tests/data/source/test_source2.parquet",
+        "tests/tape_tests/data/object/test_object2.parquet",
+        id_col="ps1_objid",
+        time_col="midPointTai",
+        band_col="filterName",
+        flux_col="psFlux",
+        err_col="psFluxErr",
+    )
+
     if func_label == "mean":
 
         def my_mean(flux):