diff --git a/tests/conftest.py b/tests/conftest.py
index 7f222011..13a27062 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -14,6 +14,30 @@
 
 import pytest
 
+#: Trace patterns that ``pytest-rerunfailures`` retries on the live-API test
+#: modules: a transient upstream 429/5xx or dropped connection is retried,
+#: deterministic failures (assertion errors, 4xx, etc.) are not. The OGC engine
+#: renders a status error as ``"<status>: ..."`` while the legacy ``query`` path
+#: renders ``"HTTP <status> ..."``, so the status pattern allows either shape;
+#: the chunked fan-out wraps a transient sub-request as ``QuotaExhausted`` /
+#: ``ServiceInterrupted``.
+_TRANSIENT_RERUN_PATTERNS = [
+    r"(?:RateLimited|ServiceUnavailable|RuntimeError):\s*(?:HTTP\s+)?(?:429|5\d\d)",
+    r"(?:QuotaExhausted|ServiceInterrupted):",
+    r"Connect(ion)?Error",  # requests' ConnectionError + httpx' ConnectError
+    r"ReadTimeout|ConnectTimeout|Timeout",
+]
+
+#: Apply to a test module (``pytestmark = flaky_api``) or class (``@flaky_api``)
+#: that hits live USGS services, so a transient upstream failure is retried
+#: instead of failing CI. Mocked tests are unaffected — the patterns match only
+#: real round-trip error traces.
+flaky_api = pytest.mark.flaky(
+    reruns=2,
+    reruns_delay=5,
+    only_rerun=_TRANSIENT_RERUN_PATTERNS,
+)
+
 
 def pytest_collection_modifyitems(config, items):
     """Apply relaxed ``pytest-httpx`` strict-mode settings to every test
diff --git a/tests/ngwmn_test.py b/tests/ngwmn_test.py
index 77b73d6e..ac9b191b 100644
--- a/tests/ngwmn_test.py
+++ b/tests/ngwmn_test.py
@@ -16,16 +16,11 @@
 
 from dataretrieval import ngwmn
 from dataretrieval.utils import BaseMetadata
+from tests.conftest import flaky_api
 
-pytestmark = pytest.mark.flaky(
-    reruns=2,
-    reruns_delay=5,
-    only_rerun=[
-        r"(?:RateLimited|RuntimeError):\s*(?:429|5\d\d):",
-        r"Connect(ion)?Error",
-        r"ReadTimeout|ConnectTimeout|Timeout",
-    ],
-)
+# These hit the live NGWMN OGC API, so retry transient upstream failures
+# rather than flaking CI (see ``conftest.flaky_api``).
+pytestmark = flaky_api
 
 # A site with water-level, construction, and lithology records (per the R
 # dataRetrieval NGWMN examples), plus a non-USGS-agency id to exercise the
diff --git a/tests/nwis_test.py b/tests/nwis_test.py
index dab46d6a..5eb379cd 100644
--- a/tests/nwis_test.py
+++ b/tests/nwis_test.py
@@ -24,6 +24,7 @@
     preformat_peaks_response,
     what_sites,
 )
+from tests.conftest import flaky_api
 
 START_DATE = "2018-01-24"
 END_DATE = "2018-01-25"
@@ -31,6 +32,10 @@
 DATETIME_COL = "datetime"
 SITENO_COL = "site_no"
 
+# Several tests in this module hit the live NWIS services, so retry a transient
+# upstream failure rather than failing CI (see ``conftest.flaky_api``).
+pytestmark = flaky_api
+
 
 def _load_mock_json(file_name):
     """Helper to load mock JSON from tests/data."""
@@ -254,21 +259,27 @@ def test_get_record_defunct_service_water_use(self):
 class TestTZ:
     """Tests relating to GitHub Issue #60."""
 
-    sites, _ = what_sites(stateCd="MD")
+    @pytest.fixture(scope="class")
+    def sites(self):
+        # Fetch once per class, at test time (not at collection) so a transient
+        # upstream failure is retried by the module ``flaky`` marker instead of
+        # aborting collection — a class-body call cannot be reran.
+        sites, _ = what_sites(stateCd="MD")
+        return sites
 
-    def test_multiple_tz_01(self):
+    def test_multiple_tz_01(self, sites):
         """Test based on GitHub Issue #60 - error merging different time zones."""
         # this test fails before issue #60 is fixed
-        iv, _ = get_iv(sites=self.sites.site_no.values[:25].tolist())
+        iv, _ = get_iv(sites=sites.site_no.values[:25].tolist())
         # assert that the datetime column exists
         assert "datetime" in iv.index.names
         # assert that it is a datetime type
         assert isinstance(iv.index[0][1], datetime.datetime)
 
-    def test_multiple_tz_02(self):
+    def test_multiple_tz_02(self, sites):
         """Test based on GitHub Issue #60 - confirm behavior for same tz."""
         # this test passes before issue #60 is fixed
-        iv, _ = get_iv(sites=self.sites.site_no.values[:20].tolist())
+        iv, _ = get_iv(sites=sites.site_no.values[:20].tolist())
         # assert that the datetime column exists
         assert "datetime" in iv.index.names
         # assert that it is a datetime type
diff --git a/tests/utils_test.py b/tests/utils_test.py
index abff0974..77b090ae 100644
--- a/tests/utils_test.py
+++ b/tests/utils_test.py
@@ -6,8 +6,12 @@
 import pytest
 
 from dataretrieval import exceptions, nwis, utils
+from tests.conftest import flaky_api
 
 
+# Test_query hits the live NWIS services (the rest of this module is mocked),
+# so retry transient upstream failures rather than flaking CI.
+@flaky_api
 class Test_query:
     """Tests of the query function."""
 
diff --git a/tests/waterdata_test.py b/tests/waterdata_test.py
index b48859c4..92b68618 100644
--- a/tests/waterdata_test.py
+++ b/tests/waterdata_test.py
@@ -40,33 +40,12 @@
     _get_args,
     _normalize_str_iterable,
 )
+from tests.conftest import flaky_api
 
-# Most tests in this module call the live USGS Water Data API. After
-# PR #273, transient upstream errors (5xx / 429 / connection drops)
-# propagate instead of silently truncating, which makes CI susceptible
-# to flaking on a brief upstream blip. Auto-retry such failures, but
-# only for the narrow set of transient-error trace patterns below —
-# library bugs raising other exception types still fail on the first
-# try. The marker is attached to every test in the module, but the
-# patterns match only traces produced by real network round-trips
-# (``_raise_for_non_200`` output, ``requests`` exceptions), so tests
-# using ``httpx_mock`` or ``mock.patch`` are no-ops for the rerun.
-pytestmark = pytest.mark.flaky(
-    reruns=2,
-    reruns_delay=5,
-    only_rerun=[
-        # Transient HTTP errors (429 / 5xx) on the direct path: RateLimited /
-        # ServiceUnavailable carry a "<status>: ..." message (the RuntimeError
-        # shape is kept for any legacy call site).
-        r"(?:RateLimited|ServiceUnavailable|RuntimeError):\s*(?:429|5\d\d):",
-        # The chunked fan-out wraps a transient sub-request as a ChunkInterrupted
-        # subclass (QuotaExhausted for 429, ServiceInterrupted for 5xx), whose
-        # message has no leading status token.
-        r"(?:QuotaExhausted|ServiceInterrupted):",
-        r"Connect(ion)?Error",  # requests' ConnectionError + httpx' ConnectError
-        r"ReadTimeout|ConnectTimeout|Timeout",
-    ],
-)
+# Most tests in this module call the live USGS Water Data API; transient
+# upstream errors propagate (since #273) instead of silently truncating, so
+# retry them rather than flaking CI (see ``conftest.flaky_api``).
+pytestmark = flaky_api
 
 
 @pytest.fixture(autouse=True)