diff --git a/Tests/helper.py b/Tests/helper.py
index 924b9733403..31bd2abe76a 100644
--- a/Tests/helper.py
+++ b/Tests/helper.py
@@ -207,7 +207,14 @@ def mark_if_feature_version(
     return pytest.mark.pil_noop_mark()
 
 
+def is_pypy() -> bool:
+    return sys.implementation.name == "pypy"
+
+
 @pytest.mark.skipif(sys.platform.startswith("win32"), reason="Requires Unix or macOS")
+# Per https://stackoverflow.com/a/29007723/51685, due to JIT compilation,
+# RSS utilization is known to grow so measuring it doesn't make that much sense.
+@pytest.mark.skipif(is_pypy(), reason="max RSS utilization is not stable on PyPy")
 class PillowLeakTestCase:
     # requires unix/macOS
     iterations = 100  # count
@@ -332,7 +339,3 @@ def is_ppc64le() -> bool:
 
 def is_win32() -> bool:
     return sys.platform.startswith("win32")
-
-
-def is_pypy() -> bool:
-    return sys.implementation.name == "pypy"
diff --git a/Tests/test_image.py b/Tests/test_image.py
index 24a61ab5b4d..d0516b849f2 100644
--- a/Tests/test_image.py
+++ b/Tests/test_image.py
@@ -572,7 +572,9 @@ def test_check_size(self) -> None:
         assert isinstance(i.size, tuple)
 
     @timeout_unless_slower_valgrind(0.75)
-    @pytest.mark.parametrize("size", ((0, 100000000), (100000000, 0)))
+    @pytest.mark.parametrize(
+        "size", ((0, 10_000_000), (10_000_000, 0)), ids=("tall", "wide")
+    )
     def test_empty_image(self, size: tuple[int, int]) -> None:
         Image.new("RGB", size)