unexpected timestamps behavior

[bordingj]

If I run

pd.to_datetime('now') == pd.to_datetime('now', utc=True)

or

pd.Timestamp('now', tz='utc') == pd.Timestamp('now')

I get

TypeError: Cannot compare tz-naive and tz-aware timestamps

which makes sense.

However if I compare a tz-naive DatetimeIndex og Series with a tz-aware timestamp I dont get an error

In[1]: date
Out[1]: Timestamp('2016-01-05 00:00:00+0000', tz='UTC')

In[2]: pd.to_datetime(['2012-01-06T00:00:00.000000000', '2013-01-08T00:00:00.000000000',
                '2014-01-07T00:00:00.000000000', '2015-01-06T00:00:00.000000000',
                '2016-01-05T00:00:00.000000000']) == date
Out[2]: array([False, False, False, False,  True])

In[3]: pd.Series(pd.to_datetime(['2012-01-06T00:00:00.000000000', '2013-01-08T00:00:00.000000000',
                '2014-01-07T00:00:00.000000000', '2015-01-06T00:00:00.000000000',
                '2016-01-05T00:00:00.000000000'])) == date
Out[3]: 
0    False
1    False
2    False
3    False
4     True
dtype: bool

and if I compare a datetime64[ns] numpy array to the tz-aware timestamp I also dont get an error, plus I get wrong results:

In[4]: pd.to_datetime(['2012-01-06T00:00:00.000000000', '2013-01-08T00:00:00.000000000',
                '2014-01-07T00:00:00.000000000', '2015-01-06T00:00:00.000000000',
                '2016-01-05T00:00:00.000000000']).values == date
Out[4]: array([False, False, False, False, False])

I am running:
pandas: 0.22.0
numpy: 1.14.0
pytz: 2017.3
python: 3.6.3

on ubuntu 16.04

[TomAugspurger]

There's been a lot of work on this recently. Comparing a tz-naive and tz-aware datetimeindex now raises on master: #18376