diff --git a/.github/workflows/gpu_tests.yml b/.github/workflows/gpu_tests.yml
index b73f85affeb..5094075dafd 100644
--- a/.github/workflows/gpu_tests.yml
+++ b/.github/workflows/gpu_tests.yml
@@ -96,6 +96,9 @@ jobs:
           COVERAGE_FILE: ${{ github.workspace }}/.coverage
           # Explicit: HOME is /github/home in container jobs, not the /root the cache steps use
           MODELOPT_WHEELHOUSE: /root/.cache/wheelhouse
+          # sys.monitoring records the same lines as the default C tracer at a fraction of the
+          # overhead on import-heavy worker processes. Needs Python 3.12+.
+          COVERAGE_CORE: sysmon
         run: |
           # Use `python3` (the vllm image has no `python` on PATH)
           python3 -m pip install nox && nox -s ${{ matrix.example }}
diff --git a/.github/workflows/regression_tests.yml b/.github/workflows/regression_tests.yml
index 9c1154a2b65..dd41c3993b9 100644
--- a/.github/workflows/regression_tests.yml
+++ b/.github/workflows/regression_tests.yml
@@ -60,6 +60,9 @@ jobs:
         env:
           COVERAGE_PROCESS_START: ${{ github.workspace }}/pyproject.toml
           COVERAGE_FILE: ${{ github.workspace }}/.coverage
+          # sys.monitoring records the same lines as the default C tracer at a fraction of the
+          # overhead on import-heavy worker processes. Needs Python 3.12+.
+          COVERAGE_CORE: sysmon
         run: python -m pip install nox && nox -s regression
       - name: Upload regression coverage to Codecov
         uses: codecov/codecov-action@v7
diff --git a/.github/workflows/unit_tests.yml b/.github/workflows/unit_tests.yml
index c3bae2e41be..3442e6147c1 100644
--- a/.github/workflows/unit_tests.yml
+++ b/.github/workflows/unit_tests.yml
@@ -69,6 +69,8 @@ jobs:
           COVERAGE_FILE: ${{ github.workspace }}/.coverage
           # Everything but PRs also runs the slow tests marked ``manual`` (e.g. AutoNAS).
           PYTEST_EXTRA: ${{ github.event_name != 'pull_request' && '--run-manual' || '' }}
+          # sys.monitoring records the same lines as the default C tracer at a fraction of the overhead
+          COVERAGE_CORE: sysmon
           # One single-threaded xdist worker per vCPU (`-n auto` counts physical cores); torch's
           # default threads per worker oversubscribe the runner and run slower than serial.
           PYTEST_ADDOPTS: -n 4 --dist loadfile