diff --git a/CHANGES.md b/CHANGES.md
index 09954f2b738..3a4d8fdf17e 100644
--- a/CHANGES.md
+++ b/CHANGES.md
@@ -19,6 +19,8 @@
 
 - Single-character closing docstring quotes are no longer moved to their own line as
   this is invalid. This was a bug introduced in version 22.6.0. (#3166)
+- `--skip-string-normalization` / `-S` now prevents docstring prefixes from being
+  normalized as expected (#3168)
 
 ### _Blackd_
 
diff --git a/src/black/linegen.py b/src/black/linegen.py
index 20f3ac6fffb..1f132b7888f 100644
--- a/src/black/linegen.py
+++ b/src/black/linegen.py
@@ -293,7 +293,24 @@ def visit_STRING(self, leaf: Leaf) -> Iterator[Line]:
         if is_docstring(leaf) and "\\\n" not in leaf.value:
             # We're ignoring docstrings with backslash newline escapes because changing
             # indentation of those changes the AST representation of the code.
-            docstring = normalize_string_prefix(leaf.value)
+            if Preview.normalize_docstring_quotes_and_prefixes_properly in self.mode:
+                # There was a bug where --skip-string-normalization wouldn't stop us
+                # from normalizing docstring prefixes. To maintain stability, we can
+                # only address this buggy behaviour while the preview style is enabled.
+                if self.mode.string_normalization:
+                    docstring = normalize_string_prefix(leaf.value)
+                    # visit_default() does handle string normalization for us, but
+                    # since this method acts differently depending on quote style (ex.
+                    # see padding logic below), there's a possibility for unstable
+                    # formatting as visit_default() is called *after*. To avoid a
+                    # situation where this function formats a docstring differently on
+                    # the second pass, normalize it early.
+                    docstring = normalize_string_quotes(docstring)
+                else:
+                    docstring = leaf.value
+            else:
+                # ... otherwise, we'll keep the buggy behaviour >.<
+                docstring = normalize_string_prefix(leaf.value)
             prefix = get_string_prefix(docstring)
             docstring = docstring[len(prefix) :]  # Remove the prefix
             quote_char = docstring[0]
diff --git a/src/black/mode.py b/src/black/mode.py
index 896c516df79..b7359fab213 100644
--- a/src/black/mode.py
+++ b/src/black/mode.py
@@ -145,12 +145,13 @@ def supports_feature(target_versions: Set[TargetVersion], feature: Feature) -> b
 class Preview(Enum):
     """Individual preview style features."""
 
-    string_processing = auto()
-    remove_redundant_parens = auto()
-    one_element_subscript = auto()
     annotation_parens = auto()
     long_docstring_quotes_on_newline = auto()
+    normalize_docstring_quotes_and_prefixes_properly = auto()
+    one_element_subscript = auto()
     remove_block_trailing_newline = auto()
+    remove_redundant_parens = auto()
+    string_processing = auto()
 
 
 class Deprecated(UserWarning):
diff --git a/tests/data/miscellaneous/docstring_preview_no_string_normalization.py b/tests/data/miscellaneous/docstring_preview_no_string_normalization.py
new file mode 100644
index 00000000000..0957231eb9c
--- /dev/null
+++ b/tests/data/miscellaneous/docstring_preview_no_string_normalization.py
@@ -0,0 +1,10 @@
+def do_not_touch_this_prefix():
+    R"""There was a bug where docstring prefixes would be normalized even with -S."""
+
+
+def do_not_touch_this_prefix2():
+    F'There was a bug where docstring prefixes would be normalized even with -S.'
+
+
+def do_not_touch_this_prefix3():
+    uR'''There was a bug where docstring prefixes would be normalized even with -S.'''
diff --git a/tests/data/simple_cases/docstring.py b/tests/data/simple_cases/docstring.py
index 7153be468c1..f08bba575fe 100644
--- a/tests/data/simple_cases/docstring.py
+++ b/tests/data/simple_cases/docstring.py
@@ -209,6 +209,13 @@ def multiline_docstring_at_line_limit():
     second line----------------------------------------------------------------------"""
 
 
+def stable_quote_normalization_with_immediate_inner_single_quote(self):
+    ''''<text here>
+
+    <text here, since without another non-empty line black is stable>
+    '''
+
+
 # output
 
 class MyClass:
@@ -417,3 +424,10 @@ def multiline_docstring_at_line_limit():
     """first line-----------------------------------------------------------------------
 
     second line----------------------------------------------------------------------"""
+
+
+def stable_quote_normalization_with_immediate_inner_single_quote(self):
+    """'<text here>
+
+    <text here, since without another non-empty line black is stable>
+    """
diff --git a/tests/test_format.py b/tests/test_format.py
index 0e1059c61e4..86339f24b86 100644
--- a/tests/test_format.py
+++ b/tests/test_format.py
@@ -139,6 +139,18 @@ def test_docstring_no_string_normalization() -> None:
     assert_format(source, expected, mode)
 
 
+def test_preview_docstring_no_string_normalization() -> None:
+    """
+    Like test_docstring but with string normalization off *and* the preview style
+    enabled.
+    """
+    source, expected = read_data(
+        "miscellaneous", "docstring_preview_no_string_normalization"
+    )
+    mode = replace(DEFAULT_MODE, string_normalization=False, preview=True)
+    assert_format(source, expected, mode)
+
+
 def test_long_strings_flag_disabled() -> None:
     """Tests for turning off the string processing logic."""
     source, expected = read_data("miscellaneous", "long_strings_flag_disabled")