diff --git a/torch/_dispatch/python.py b/torch/_dispatch/python.py
index f0814889ba2d..bbc4c277273e 100644
--- a/torch/_dispatch/python.py
+++ b/torch/_dispatch/python.py
@@ -4,6 +4,8 @@ import unittest.mock
 import torch
 import torch.utils._pytree as pytree
 import itertools
+from typing import Iterator
+import torch._ops
 
 __all__ = ['enable_python_dispatcher', 'no_python_dispatcher']
 
@@ -25,7 +27,24 @@ def enable_python_dispatcher():
 
 CROSSREF_FUNCTIONALIZE = False
 
-def all_known_overloads():
+def all_py_loaded_overloads() -> Iterator[torch._ops.OpOverload]:
+    """
+    Warning: the set of overloads this will report is very subtle.  It is precisely
+    the set of torch.ops functions that have actually been accessed from Python
+    (e.g., we actually called torch.ops.aten.blah at some point.  This is DIFFERENT
+    from the set of registered operators, which will in general be a larger set,
+    as this would include all operators which we ran C++ static initializers or
+    Python operator registration on.  This does not eagerly populate the list on
+    torch.ops.aten; this list is lazy!
+
+    In other words, this is good for traversing over everything that has an
+    OpOverload object allocated in Python.  We use it for cache invalidation, but
+    don't rely on this list being complete.
+
+    Note that even if we did report all C++ registered overloads, this isn't guaranteed
+    to be complete either, as a subsequent lazy load of a library which triggers more
+    registrations could add more things to the set.
+    """
     for ns in torch.ops:
         packets = getattr(torch.ops, ns)
         for op_name in packets:
@@ -131,12 +150,12 @@ def make_crossref_functionalize(op, final_key):
 # for debugging purposes.
 @contextmanager
 def enable_crossref_functionalize():
-    for op in all_known_overloads():
+    for op in all_py_loaded_overloads():
         op._uncache_dispatch(torch._C.DispatchKey.Functionalize)
     try:
         with enable_python_dispatcher(), unittest.mock.patch(
                 'torch._dispatch.python.CROSSREF_FUNCTIONALIZE', True):
             yield
     finally:
-        for op in all_known_overloads():
+        for op in all_py_loaded_overloads():
             op._uncache_dispatch(torch._C.DispatchKey.Functionalize)