diff --git a/CHANGES.md b/CHANGES.md
index 36cbd0039e0..eefcc8cfa71 100644
--- a/CHANGES.md
+++ b/CHANGES.md
@@ -22,6 +22,7 @@
 - Fix `string_processing` crashing on unassigned long string literals with trailing
   commas (one-item tuples) (#4929)
 - Simplify implementation of the power operator "hugging" logic (#4918)
+- Always wrap `in` clauses of comprehensions if they have delimiters (#4881)
 
 ### Configuration
 
diff --git a/src/black/brackets.py b/src/black/brackets.py
index 44a3c9a2946..f029dec7c7e 100644
--- a/src/black/brackets.py
+++ b/src/black/brackets.py
@@ -340,8 +340,14 @@ def max_delimiter_priority_in_atom(node: LN) -> Priority:
     if not (first.type == token.LPAR and last.type == token.RPAR):
         return 0
 
+    return max_delimiter_priority(node.children[1:-1])
+
+
+def max_delimiter_priority(children: list[LN]) -> Priority:
+    """Return maximum delimiter priority in a list of leaves and nodes."""
+
     bt = BracketTracker()
-    for c in node.children[1:-1]:
+    for c in children:
         if isinstance(c, Leaf):
             bt.mark(c)
         else:
diff --git a/src/black/linegen.py b/src/black/linegen.py
index f907070fe25..6cb68c863e4 100644
--- a/src/black/linegen.py
+++ b/src/black/linegen.py
@@ -15,6 +15,7 @@
     DOT_PRIORITY,
     STRING_PRIORITY,
     get_leaves_inside_matching_brackets,
+    max_delimiter_priority,
     max_delimiter_priority_in_atom,
 )
 from black.comments import (
@@ -639,9 +640,43 @@ def visit_tstring(self, node: Node) -> Iterator[Line]:
 
     def visit_comp_for(self, node: Node) -> Iterator[Line]:
         if Preview.wrap_comprehension_in in self.mode:
-            normalize_invisible_parens(
-                node, parens_after={"in"}, mode=self.mode, features=self.features
-            )
+            check_lpar = False
+            for child in node.children:
+                if not check_lpar:
+                    check_lpar = isinstance(child, Leaf) and child.value == "in"
+                    continue
+
+                check_lpar = isinstance(child, Leaf) and child.value == "in"
+
+                is_wrapped = (
+                    len(child.children) == 3
+                    and is_lpar_token(child.children[0])
+                    and is_rpar_token(child.children[-1])
+                )
+                max_delimiter = max_delimiter_priority(
+                    (child.children[1] if is_wrapped else child).children
+                )
+
+                if child.type == syms.atom:
+                    if is_wrapped and child.children[1].type == syms.test:
+                        continue
+                    if max_delimiter > DOT_PRIORITY:
+                        if not is_wrapped and maybe_make_parens_invisible_in_atom(
+                            child,
+                            parent=node,
+                            mode=self.mode,
+                            features=self.features,
+                        ):
+                            wrap_in_parentheses(node, child, visible=True)
+                    elif maybe_make_parens_invisible_in_atom(
+                        child, parent=node, mode=self.mode, features=self.features
+                    ):
+                        wrap_in_parentheses(node, child, visible=False)
+                else:
+                    wrap_in_parentheses(
+                        node, child, visible=max_delimiter > DOT_PRIORITY
+                    )
+
         yield from self.visit_default(node)
 
     def visit_old_comp_for(self, node: Node) -> Iterator[Line]:
diff --git a/tests/data/cases/preview_wrap_comprehension_in.py b/tests/data/cases/preview_wrap_comprehension_in.py
index e457f0e772f..2d51a269f3e 100644
--- a/tests/data/cases/preview_wrap_comprehension_in.py
+++ b/tests/data/cases/preview_wrap_comprehension_in.py
@@ -70,6 +70,14 @@
     )
 }
 
+# Short `in`s with delimiters
+filtered_trait_instructions: list[str] = [
+    trait for trait in (traits_instructions or []) if trait
+]
+filtered_trait_instructions: list[str] = [
+    trait for trait in traits_instructions or [] if trait
+]
+
 # output
 [
     a
@@ -159,3 +167,11 @@
     key_with_super_really_long_name: key_with_super_really_long_name
     for key in dictionary
 }
+
+# Short `in`s with delimiters
+filtered_trait_instructions: list[str] = [
+    trait for trait in (traits_instructions or []) if trait
+]
+filtered_trait_instructions: list[str] = [
+    trait for trait in (traits_instructions or []) if trait
+]
diff --git a/tests/optional.py b/tests/optional.py
index f9bceb6f9ff..84960481cbe 100644
--- a/tests/optional.py
+++ b/tests/optional.py
@@ -91,7 +91,7 @@ def pytest_configure(config: "Config") -> None:
     passed_args = config.getoption("run_optional")
     if passed_args:
         ot_run.update(itertools.chain.from_iterable(a.split(",") for a in passed_args))
-    ot_run |= {no(excluded) for excluded in ot_markers - ot_run}
+    ot_run |= {no(excluded) for excluded in (ot_markers - ot_run)}
     ot_markers |= {no(m) for m in ot_markers}
 
     log.info("optional tests to run: %s", ot_run)
diff --git a/tests/test_docs.py b/tests/test_docs.py
index e09124acb2e..271905260c9 100644
--- a/tests/test_docs.py
+++ b/tests/test_docs.py
@@ -64,7 +64,7 @@ def test_feature_lists_are_up_to_date() -> None:
         future_style = f.readlines()
     preview_error = check_feature_list(
         future_style,
-        {feature.name for feature in set(Preview) - UNSTABLE_FEATURES},
+        {feature.name for feature in (set(Preview) - UNSTABLE_FEATURES)},
         "preview",
     )
     assert preview_error is None, preview_error