Crash when executing function which filters numpy array

Using

library(reticulate)
source_python(pythonfile.py)

to source pythonfile.py wit the content

import numpy as np

def testfun(idx):
    idx = int(idx)
    # Some MT numbers and corresponding TALYS file names #
    mts = np.array([1, 2, 3])
    mt2name = np.array(['a', 'b', 'c'])
    if idx in mts:
        ret = mt2name[idx == mts][0]
        return ret
    return 'undefined'

and then running from R

testfun(1L)

leads to a crash of R with the error message

*** caught segfault ***
address 0x8, cause 'memory not mapped'

Additional information:
R version 3.4.4 (2018-03-15) -- "Someone to Lean On"
Python 2.7.15 |Anaconda, Inc.| (default, May 1 2018, 23:32:55)

Using the function testfun(1) directly from the python shell works without problems.

Start by reproducing the crash with pythonfile.py, testfun, and the R call testfun(1L), then compare it with calling the function directly from Python. Trace the reticulate boundary involved in the NumPy filtering expression. Done means the R process no longer segfaults and the function returns its expected value.