add note about caching granularity from PR comments

src/python/pants/backend/python/tasks/python_tool_prep_base.py

@@ -106,6 +106,11 @@ def _build_tool_pex(self, tool_subsystem, interpreter, pex_path):
       pex_builder.freeze()
 
   def _generate_fingerprinted_pex_path(self, tool_subsystem, interpreter):
+    # `tool_subsystem.get_requirement_specs()` is a list, but order shouldn't actually matter. This
+    # should probably be sorted, but it's possible a user could intentionally tweak order to work
+    # around a particular requirement resolution resolve-order issue. In practice the lists are
+    # expected to be mostly static, so we accept the risk of too-fine-grained caching creating lots
+    # of pexes in the cache dir.
     specs_fingerprint = stable_json_sha1(tool_subsystem.get_requirement_specs())
     return os.path.join(
       get_pants_cachedir(),

tests/python/pants_test/backend/python/tasks/test_python_tool.py

@@ -42,10 +42,6 @@ def prepare(cls, options, round_manager):
     super(ToolTask, cls).prepare(options, round_manager)
     round_manager.require_data(ToolPrep.tool_instance_cls)
 
-  @classmethod
-  def product_types(cls):
-    return ['test_pex']
-
   def execute(self):
     tool_for_pex = self.context.products.get_data(ToolPrep.tool_instance_cls)
     stdout, stderr, exit_code, _ = tool_for_pex.output(['--version'])