BUG: df.groupby().apply() behaves inconsistently when there's only one group

[ypsah]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

import pandas as pd

PLAYERS = pd.DataFrame({
    "name": ["alice", "bob", "carol"],
    "team": "blue",
    "score": [10, 5, 12],
})

def _has_won(teammates: pd.DataFrame) -> pd.Series:
    """
    To win, one must score at least 6 and their team's average score must be >= 6
    """
    return (teammates["score"].mean() >= 6) & (teammates["score"] >= 6)

HAS_WON = PLAYERS.groupby(["team"], group_keys=False).apply(_has_won)
print(HAS_WON)

Issue Description

In the snippet above, HAS_WON has the wrong shape:

score     0      1     2
team                    
blue   True  False  True

This only happens when there's a single group in the dataframe.
Using:

PLAYERS = pd.DataFrame({
    "name": ["alice", "bob", "carol"],
    "team": ["blue", "red", "blue"],
    "score": [10, 5, 12],
})

The output becomes:

0     True
2     True
1    False
Name: score, dtype: bool

Expected Behavior

I would expect the reproducer I sent to produce a series whose shape does not depend on the number of groups.

At the moment, I work around the issue as follows:

teams = PLAYERS["team"].unique()
if len(teams) == 1:
    HAS_WON = _has_won(PLAYERS)
else:
    HAS_WON = PLAYERS.groupby(["team"], group_keys=False).apply(_has_won)

Installed Versions

INSTALLED VERSIONS

commit : 0f43794
python : 3.11.3.final.0
python-bits : 64
OS : Linux
OS-release : 6.4.11-arch2-1
Version : #1 SMP PREEMPT_DYNAMIC Sat, 19 Aug 2023 15:38:34 +0000
machine : x86_64
processor :
byteorder : little
LC_ALL : None
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 2.0.3
numpy : 1.25.2
pytz : 2023.3
dateutil : 2.8.2
setuptools : 68.0.0
pip : 23.1.2
Cython : None
pytest : None
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : None
IPython : None
pandas_datareader: None
bs4 : None
bottleneck : None
brotli : None
fastparquet : None
fsspec : None
gcsfs : None
matplotlib : None
numba : None
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pyreadstat : None
pyxlsb : None
s3fs : None
scipy : None
snappy : None
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
zstandard : None
tzdata : 2023.3
qtpy : None
pyqt5 : None

[ypsah]

This feels related to #44803.

[rhshadrach]

Duplicate of #42749. This particular operation should be performed without apply (breakup the operation and use .transform("mean") instead) - not only will you get better behavior, but it will be much faster too.