diff --git a/beets/library.py b/beets/library.py
index 006630c08..10e54b2a1 100644
--- a/beets/library.py
+++ b/beets/library.py
@@ -790,26 +790,19 @@ class Item(LibModel):
         if beets.config['asciify_paths']:
             subpath = unidecode(subpath)
 
-        # Truncate components and remove forbidden characters.
-        subpath = util.sanitize_path(subpath, self._db.replacements)
-
-        # Encode for the filesystem.
-        if not fragment:
-            subpath = bytestring_path(subpath)
-
-        # Preserve extension.
-        _, extension = os.path.splitext(self.path)
-        if fragment:
-            # Outputting Unicode.
-            extension = extension.decode('utf8', 'ignore')
-        subpath += extension.lower()
-
-        # Truncate too-long components.
         maxlen = beets.config['max_filename_length'].get(int)
         if not maxlen:
             # When zero, try to determine from filesystem.
             maxlen = util.max_filename_length(self._db.directory)
-        subpath = util.truncate_path(subpath, maxlen)
+
+        subpath, fellback = util.legalize_path(
+            subpath, self._db.replacements, maxlen,
+            os.path.splitext(self.path)[1], fragment
+        )
+
+        # Print an error message if legalize fell back to default replacements
+        if fellback:
+            log.warning(u'fell back to default replacements when naming file')
 
         if fragment:
             return subpath
diff --git a/beets/util/__init__.py b/beets/util/__init__.py
index 68be740f6..42971fdf6 100644
--- a/beets/util/__init__.py
+++ b/beets/util/__init__.py
@@ -545,6 +545,63 @@ def truncate_path(path, length=MAX_FILENAME_LENGTH):
     return os.path.join(*out)
 
 
+def _legalize_stage(path, replacements, length, extension, fragment):
+    """ Performs sanitization of the path, then encodes for the filesystem,
+    appends the extension and truncates. Returns the path (unicode if fragment
+    is set, otherwise bytes), and whether truncation was required.
+    """
+    # Perform an initial sanitization including user replacements.
+    path = sanitize_path(path, replacements)
+
+    # Encode for the filesystem.
+    if not fragment:
+        path = bytestring_path(path)
+
+    # Preserve extension.
+    path += extension.lower()
+
+    # Truncate too-long components.
+    pre_truncate_path = path
+    path = truncate_path(path, length)
+
+    return path, path != pre_truncate_path
+
+
+def legalize_path(path, replacements, length, extension, fragment):
+    """ Perform up to three calls to _legalize_stage, to generate a stable
+    output path, taking user replacements into consideration if possible. The
+    limited number of iterations avoids the possibility of an infinite loop of
+    sanitization and truncation operations, which could be caused by some user
+    replacements.
+    """
+
+    if fragment:
+        # Outputting Unicode.
+        extension = extension.decode('utf8', 'ignore')
+
+    first_stage_path, _ = _legalize_stage(
+        path, replacements, length, extension, fragment
+    )
+
+    # Convert back to Unicode with extension removed.
+    first_stage_path = os.path.splitext(displayable_path(first_stage_path))[0]
+
+    # Re-sanitize following truncation (including user replacements).
+    second_stage_path, retruncated = _legalize_stage(
+        first_stage_path, replacements, length, extension, fragment
+    )
+
+    # If the path was once again truncated, discard user replacements.
+    if retruncated:
+        second_stage_path, _ = _legalize_stage(
+            first_stage_path, None, length, extension, fragment
+        )
+
+        return second_stage_path, True
+    else:
+        return second_stage_path, False
+
+
 def str2bool(value):
     """Returns a boolean reflecting a human-entered string."""
     return value.lower() in ('yes', '1', 'true', 't', 'y')
diff --git a/docs/reference/config.rst b/docs/reference/config.rst
index d737bba69..9000f070a 100644
--- a/docs/reference/config.rst
+++ b/docs/reference/config.rst
@@ -119,6 +119,11 @@ compatibility with Windows-influenced network filesystems like Samba).
 Trailing dots and trailing whitespace, which can cause problems on Windows
 clients, are also removed.
 
+When replacements other than the defaults are used, it is possible that they
+will increase the length of the path. In the scenario where this leads to a
+conflict with the maximum filename length, the default replacements will be
+used to resolve the conflict and beets will display a warning.
+
 Note that paths might contain special characters such as typographical
 quotes (``“”``). With the configuration above, those will not be
 replaced as they don't match the typewriter quote (``"``). To also strip these
diff --git a/test/test_library.py b/test/test_library.py
index a4b4d08a8..30b0d6fc4 100644
--- a/test/test_library.py
+++ b/test/test_library.py
@@ -452,8 +452,7 @@ class DestinationTest(_common.TestCase):
         self.assertEqual(self.i.destination(),
                          np('base/one/_.mp3'))
 
-    @unittest.skip('unimplemented: #496')
-    def test_truncation_does_not_conflict_with_replacement(self):
+    def test_legalize_path_one_for_one_replacement(self):
         # Use a replacement that should always replace the last X in any
         # path component with a Z.
         self.lib.replacements = [
@@ -466,7 +465,23 @@ class DestinationTest(_common.TestCase):
 
         # The final path should reflect the replacement.
         dest = self.i.destination()
-        self.assertTrue('XZ' in dest)
+        self.assertEqual(dest[-2:], 'XZ')
+
+    def test_legalize_path_one_for_many_replacement(self):
+        # Use a replacement that should always replace the last X in any
+        # path component with four Zs.
+        self.lib.replacements = [
+            (re.compile(r'X$'), u'ZZZZ'),
+        ]
+
+        # Construct an item whose untruncated path ends with a Y but whose
+        # truncated version ends with an X.
+        self.i.title = 'X' * 300 + 'Y'
+
+        # The final path should ignore the user replacement and create a path
+        # of the correct length, containing Xs.
+        dest = self.i.destination()
+        self.assertEqual(dest[-2:], 'XX')
 
 
 class ItemFormattedMappingTest(_common.LibTestCase):