“TypeError: data type not understood” comparing dtype np.datetime64 with issubdtype

Reproducing code example:

import pandas as pd
import numpy as np
df_log = pd.DataFrame(columns=['ts', 'event'])
ts_start = 1582582300
ts_end = ts_start + 30
df_log['ts']=pd.Series(np.arange(ts_start, ts_end))
df_log.loc[df_log['ts']==1582582300 + 3, 'event']="Entered room"
df_log.loc[df_log['ts']==1582582300 + 12, 'event']="Door opened"
df_log.loc[df_log['ts']==1582582300 + 18, 'event']="Door closed"
df_log.loc[df_log['ts']==1582582300 + 25, 'event']="Left room"
df_log['ts']=pd.to_datetime(df_log["ts"], unit='s', utc=True).dt.tz_convert('US/Central')
for i in range(len(df_log.dtypes)):
    dtype = df_log.dtypes[i]
    print("testing:", dtype)
    if np.issubdtype(dtype, np.datetime64):
         print("is subtype")

Error message:

testing: datetime64[ns, US/Central]
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-2-225a4521e525> in <module>
     13     dtype = df_log.dtypes[i]
     14     print("testing:", dtype)
---> 15     if np.issubdtype(dtype, np.datetime64):
     16          print("is subtype")

/usr/local/lib/python3.7/site-packages/numpy/core/numerictypes.py in issubdtype(arg1, arg2)
    391     """
    392     if not issubclass_(arg1, generic):
--> 393         arg1 = dtype(arg1).type
    394     if not issubclass_(arg2, generic):
    395         arg2_orig = arg2

TypeError: data type not understood

Numpy/Python version information:

print(numpy.__version__, sys.version)
1.18.1 3.7.6 (default, Dec 30 2019, 19:38:26) 
[Clang 11.0.0 (clang-1100.0.33.16)]

Start by running the reproducer with NumPy 1.18.1 and Python 3.7.6, focusing on the np.issubdtype call with pandas' timezone-aware datetime64 dtype. Inspect NumPy's dtype handling around the numerictypes.py path shown in the traceback and compare behavior with timezone-naive dtypes. Done means the reported comparison has defined, tested behavior instead of raising the shown TypeError.