diff --git a/litellm/caching/redis_cache.py b/litellm/caching/redis_cache.py
index f6223fb8d94..6ef99f5ffd1 100644
--- a/litellm/caching/redis_cache.py
+++ b/litellm/caching/redis_cache.py
@@ -143,6 +143,7 @@ _INCREMENT_IF_PRESENT_LUA: Final = (
 
 _LUA_COUNT: Final = TypeAdapter(int)
 _PLAIN_DECIMAL: Final = re.compile(r"-?[0-9]+(?:\.[0-9]+)?(?:[eE][-+]?[0-9]+)?")
+_PIPELINE_COMMAND_NUMBER: Final = re.compile(r"Command # ([0-9]+) \(")
 _LUA_OPTIONAL_TOTAL: Final[TypeAdapter[float | None]] = TypeAdapter(float | None)
 _OPTIONAL_COUNTS: Final = TypeAdapter(tuple[int | None, ...])
 
@@ -153,6 +154,13 @@ def _as_incrbyfloat_writes(value: float) -> str:
     return "0" if text == "-0" else text
 
 
+def _failed_after_the_first_command(error: Exception) -> bool:
+    """Whether redis-py names a pipeline command after the first as the one that failed, so the first one applied."""
+    message: Final = error.args[0] if error.args else None
+    number: Final = _PIPELINE_COMMAND_NUMBER.match(message) if isinstance(message, str) else None
+    return number is not None and int(number.group(1)) > 1
+
+
 def _plain_float(reply: object) -> float | None:
     """The finite float a stored value spells in plain decimal or exponent form, or None for anything else."""
     text: Final = reply.decode(errors="replace") if isinstance(reply, bytes) else reply
@@ -1494,8 +1502,9 @@ class RedisCache(BaseCache):
 
     async def _set_if_missing_and_get(self, key: str, value: float, ttl: int | None) -> tuple[bool, float] | None:
         """SET NX GET and its TTL command in one round trip: whether the key was created and the value it holds, or
-        None when the server refused a command rather than failed to answer, or the key holds a value this does not
-        read as INCRBYFLOAT would. SET NX GET is one command, so a refused one wrote nothing."""
+        None when the server refused SET NX GET rather than failed to answer, or the key holds a value this does not
+        read as INCRBYFLOAT would. SET NX GET is one command, so a refused one wrote nothing. A refused TTL command
+        raises, as it does in main's add."""
         from redis.exceptions import ClusterError, RedisError
         from redis.exceptions import ConnectionError as RedisConnectionError
         from redis.exceptions import TimeoutError as RedisTimeoutError
@@ -1510,7 +1519,8 @@ class RedisCache(BaseCache):
                     pipe.expire(key, ttl)
                 replies: Final[Sequence[object]] = await pipe.execute()  # pyright: ignore[reportUnknownVariableType, reportUnknownMemberType]  # redis generics
         except RedisError as e:
-            if isinstance(e, (RedisConnectionError, RedisTimeoutError, ClusterError)):
+            failed_to_answer: Final = isinstance(e, (RedisConnectionError, RedisTimeoutError, ClusterError))
+            if failed_to_answer or _failed_after_the_first_command(e):
                 raise
             return None
         if replies[0] is None:
diff --git a/tests/unit/caching/test_redis_cache.py b/tests/unit/caching/test_redis_cache.py
index 2a2ac5f7063..e673f16e5a6 100644
--- a/tests/unit/caching/test_redis_cache.py
+++ b/tests/unit/caching/test_redis_cache.py
@@ -1727,8 +1727,6 @@ class _KeyspacePipeline(_RaisingPipeline):
         self.commands.append(("SETNX", name, value))
 
     def _write(self, command: str) -> None:
-        if command in self.keyspace.refused:
-            raise NoPermissionError(f"this user has no permissions to run the '{command.lower()}' command")
         if self.writes_left is not None:
             if self.writes_left == 0:
                 raise ResponseError("OOM command not allowed when used memory > 'maxmemory'.")
@@ -1737,6 +1735,8 @@ class _KeyspacePipeline(_RaisingPipeline):
     def _run(self, command: tuple[object, ...]) -> object:
         name: Final = str(command[1])
         held: Final = self.keyspace.values.get(name)
+        if command[0] in self.keyspace.refused:
+            raise NoPermissionError(f"this user has no permissions to run the '{str(command[0]).lower()}' command")
         if command[0] == "EXPIRE":
             return held is not None
         self._write(str(command[0]))
@@ -1756,10 +1756,12 @@ class _KeyspacePipeline(_RaisingPipeline):
         if self.lost:
             raise RedisConnectionError("Connection closed by server.")
         replies: Final[list[object]] = []
-        for command in self.commands:
+        for number, command in enumerate(self.commands, start=1):
             try:
                 replies.append(self._run(command))
             except RedisError as e:
+                words: str = " ".join(str(part) for part in command)
+                e.args = (f"Command # {number} ({words}) of pipeline caused error: {e.args}",)
                 replies.append(e)
         first_error: Final = next((reply for reply in replies if isinstance(reply, Exception)), None)
         if first_error is not None:
@@ -1801,6 +1803,25 @@ async def test_async_increment_only_if_missing_sets_a_missing_key_once_when_redi
     added.assert_called_once_with()
 
 
+@pytest.mark.asyncio
+async def test_async_increment_only_if_missing_raises_a_refused_ttl_command_as_mains_add_does_without_adding_again(
+    monkeypatch: pytest.MonkeyPatch, redis_no_ping: object
+):
+    monkeypatch.setenv("REDIS_HOST", "https://my-test-host")
+    redis_cache: Final = RedisCache(namespace="ns")
+    keyspace: Final = _Keyspace(refused=frozenset(("EXPIRE",)))
+    added: Final = MagicMock()
+
+    with patch.object(redis_cache, "init_async_client", return_value=keyspace), pytest.raises(NoPermissionError):
+        await redis_cache.async_increment(
+            key="spend:key:k", value=6.0, ttl=60, refresh_ttl=True, only_if_missing=True, on_added=added
+        )
+
+    assert keyspace.values == {"ns:spend:key:k": "6"}
+    assert keyspace.pipelines == 1
+    added.assert_not_called()
+
+
 @pytest.mark.asyncio
 @pytest.mark.parametrize("held", ["not-a-number", "0x10", "6_0", " 6", "inf", "nan", "1e999"])
 async def test_async_increment_only_if_missing_leaves_a_value_it_cannot_read_like_incrbyfloat_to_mains_add(