diff --git a/.github/workflows/pr.yaml b/.github/workflows/pr.yaml
index 14a74618413..1dc31da8e80 100644
--- a/.github/workflows/pr.yaml
+++ b/.github/workflows/pr.yaml
@@ -164,6 +164,8 @@ jobs:
       matrix_filter: map(select(.ARCH == "amd64")) | max_by(.CUDA_VER) | [.]
       build_type: pull-request
       script: ci/cudf_pandas_scripts/pandas-tests/run.sh pr
+      # Hide test failures because they exceed the GITHUB_STEP_SUMMARY output limit.
+      test_summary_show: "none"
   #pandas-tests-diff:
   #  # diff the results of running the Pandas unit tests and publish a job summary
   #  needs: [pandas-tests-main, pandas-tests-pr]
diff --git a/ci/cudf_pandas_scripts/pandas-tests/run.sh b/ci/cudf_pandas_scripts/pandas-tests/run.sh
index be5705a9548..482af42201f 100755
--- a/ci/cudf_pandas_scripts/pandas-tests/run.sh
+++ b/ci/cudf_pandas_scripts/pandas-tests/run.sh
@@ -1,12 +1,14 @@
 #!/usr/bin/env bash
-# SPDX-FileCopyrightText: Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES.
+# SPDX-FileCopyrightText: Copyright (c) 2023-2024, NVIDIA CORPORATION & AFFILIATES.
 # All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
+set -euo pipefail
+
 PANDAS_TESTS_BRANCH=${1}
 
 rapids-logger "Running Pandas tests using $PANDAS_TESTS_BRANCH branch"
-rapids-logger "PR number: $RAPIDS_REF_NAME"
+rapids-logger "PR number: ${RAPIDS_REF_NAME:-"unknown"}"
 
 # Set the manylinux version used for downloading the wheels so that we test the
 # newer ABI wheels on the newer images that support their installation.
@@ -25,14 +27,16 @@ RAPIDS_PY_CUDA_SUFFIX="$(rapids-wheel-ctk-name-gen ${RAPIDS_CUDA_VERSION})"
 RAPIDS_PY_WHEEL_NAME="cudf_${manylinux}_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 ./local-cudf-dep
 python -m pip install $(ls ./local-cudf-dep/cudf*.whl)[test,pandas-tests]
 
-git checkout $COMMIT
+RESULTS_DIR=${RAPIDS_TESTS_DIR:-"$(mktemp -d)"}
+RAPIDS_TESTS_DIR=${RAPIDS_TESTS_DIR:-"${RESULTS_DIR}/test-results"}/
+mkdir -p "${RAPIDS_TESTS_DIR}"
 
 bash python/cudf/cudf/pandas/scripts/run-pandas-tests.sh \
   -n 10 \
   --tb=line \
-  --skip-slow \
+  -m "not slow" \
   --max-worker-restart=3 \
-  --import-mode=importlib \
+  --junitxml="${RAPIDS_TESTS_DIR}/junit-cudf-pandas.xml" \
   --report-log=${PANDAS_TESTS_BRANCH}.json 2>&1
 
 # summarize the results and save them to artifacts:
diff --git a/ci/test_wheel_cudf.sh b/ci/test_wheel_cudf.sh
index 8c42651e299..b7e8f862ed5 100755
--- a/ci/test_wheel_cudf.sh
+++ b/ci/test_wheel_cudf.sh
@@ -1,5 +1,5 @@
 #!/bin/bash
-# Copyright (c) 2023, NVIDIA CORPORATION.
+# Copyright (c) 2023-2024, NVIDIA CORPORATION.
 
 set -eou pipefail
 
@@ -22,9 +22,22 @@ RAPIDS_PY_WHEEL_NAME="cudf_${manylinux}_${RAPIDS_PY_CUDA_SUFFIX}" rapids-downloa
 # echo to expand wildcard before adding `[extra]` requires for pip
 python -m pip install $(echo ./dist/cudf*.whl)[test]
 
+RESULTS_DIR=${RAPIDS_TESTS_DIR:-"$(mktemp -d)"}
+RAPIDS_TESTS_DIR=${RAPIDS_TESTS_DIR:-"${RESULTS_DIR}/test-results"}/
+mkdir -p "${RAPIDS_TESTS_DIR}"
+
 # Run smoke tests for aarch64 pull requests
 if [[ "$(arch)" == "aarch64" && ${RAPIDS_BUILD_TYPE} == "pull-request" ]]; then
+    rapids-logger "Run smoke tests for cudf"
     python ./ci/wheel_smoke_test_cudf.py
 else
-    python -m pytest -n 8 ./python/cudf/cudf/tests
+    rapids-logger "pytest cudf"
+    pushd python/cudf/cudf/tests
+    python -m pytest \
+      --cache-clear \
+      --junitxml="${RAPIDS_TESTS_DIR}/junit-cudf.xml" \
+      --numprocesses=8 \
+      --dist=loadscope \
+      .
+    popd
 fi
diff --git a/ci/test_wheel_dask_cudf.sh b/ci/test_wheel_dask_cudf.sh
index e9162b816aa..74fcb43ddca 100755
--- a/ci/test_wheel_dask_cudf.sh
+++ b/ci/test_wheel_dask_cudf.sh
@@ -1,5 +1,5 @@
 #!/bin/bash
-# Copyright (c) 2023, NVIDIA CORPORATION.
+# Copyright (c) 2023-2024, NVIDIA CORPORATION.
 
 set -eou pipefail
 
@@ -26,5 +26,15 @@ python -m pip install --no-deps ./local-cudf-dep/cudf*.whl
 # echo to expand wildcard before adding `[extra]` requires for pip
 python -m pip install $(echo ./dist/dask_cudf*.whl)[test]
 
+RESULTS_DIR=${RAPIDS_TESTS_DIR:-"$(mktemp -d)"}
+RAPIDS_TESTS_DIR=${RAPIDS_TESTS_DIR:-"${RESULTS_DIR}/test-results"}/
+mkdir -p "${RAPIDS_TESTS_DIR}"
+
 # Run tests in dask_cudf/tests and dask_cudf/io/tests
-python -m pytest -n 8 ./python/dask_cudf/dask_cudf/
+rapids-logger "pytest dask_cudf"
+pushd python/dask_cudf/dask_cudf
+python -m pytest \
+  --junitxml="${RAPIDS_TESTS_DIR}/junit-dask-cudf.xml" \
+  --numprocesses=8 \
+  .
+popd
diff --git a/python/cudf/cudf/pandas/scripts/run-pandas-tests.sh b/python/cudf/cudf/pandas/scripts/run-pandas-tests.sh
index 4fe152cc493..319e5ba80fc 100755
--- a/python/cudf/cudf/pandas/scripts/run-pandas-tests.sh
+++ b/python/cudf/cudf/pandas/scripts/run-pandas-tests.sh
@@ -16,12 +16,13 @@
 #
 # This script creates a `pandas-testing` directory if it doesn't exist
 
+set -euo pipefail
 
 # Grab the Pandas source corresponding to the version
 # of Pandas installed.
 PANDAS_VERSION=$(python -c "import pandas; print(pandas.__version__)")
 
-PYTEST_IGNORES="--ignore=tests/io/test_user_agent.py"
+PYTEST_IGNORES="--ignore=tests/io/test_user_agent.py --ignore=tests/interchange/test_impl.py"
 
 mkdir -p pandas-testing
 cd pandas-testing
@@ -92,7 +93,7 @@ cd pandas-tests/
 # test_overwrite_warns unsafely patchs over Series.mean affecting other tests when run in parallel
 # test_complex_series_frame_alignment randomly selects a DataFrames and axis to test but particular random selection(s) always fails
 # test_numpy_ufuncs_basic compares floating point values to unbounded precision, sometimes leading to failures
-TEST_NUMPY_UFUNCS_BASIC_FLAKY="test_numpy_ufuncs_basic[float-exp] \
+TEST_NUMPY_UFUNCS_BASIC_FLAKY="not test_numpy_ufuncs_basic[float-exp] \
 and not test_numpy_ufuncs_basic[float-exp2] \
 and not test_numpy_ufuncs_basic[float-expm1] \
 and not test_numpy_ufuncs_basic[float-log] \
@@ -183,11 +184,12 @@ and not test_numpy_ufuncs_basic[nullable_float-rad2deg]"
 
 PANDAS_CI="1" python -m pytest -p cudf.pandas \
     -m "not single_cpu and not db" \
-    -k "not test_overwrite_warns and not test_complex_series_frame_alignment and not $TEST_NUMPY_UFUNCS_BASIC_FLAKY" \
+    -k "not test_overwrite_warns and not test_complex_series_frame_alignment and $TEST_NUMPY_UFUNCS_BASIC_FLAKY" \
     --durations=50 \
     --import-mode=importlib \
     -o xfail_strict=True \
-    ${PYTEST_IGNORES} $@
+    ${PYTEST_IGNORES} \
+    "$@" || [ $? = 1 ]  # Exit success if exit code was 1 (permit test failures but not other errors)
 
 mv *.json ..
 cd ..