Minimal reproduction case

import numba

# let's try to use openmp
numba.config.THREADING_LAYER = 'omp'

@numba.njit(parallel=True)
def do_stuff(a):
    for i in numba.prange(100):
        a += i
    return a


print(do_stuff(0))

When I execute that script with numba 0.40.0 I get:

Inconsistency detected by ld.so: dl-version.c: 224: _dl_check_map_versions: Assertion `needed != NULL' failed!

Environment

Linux with Python 3.7 and a virtualenv with the following lib versions:

$ pip show numba llvmlite tbb
Name: numba
Version: 0.40.0
Summary: compiling Python code using LLVM
Home-page: http://numba.github.com
Author: Anaconda, Inc.
Author-email: numba-users@continuum.io
License: BSD
Location: /opt/venvs/py37/lib/python3.7/site-packages
Requires: llvmlite, numpy
Required-by: pygbm
---
Name: llvmlite
Version: 0.25.0
Summary: lightweight wrapper around basic LLVM functionality
Home-page: http://llvmlite.pydata.org
Author: Continuum Analytics, Inc.
Author-email: numba-users@continuum.io
License: BSD
Location: /opt/venvs/py37/lib/python3.7/site-packages
Requires: 
Required-by: numba
---
Name: tbb
Version: 2019.0
Summary: High level abstract threading library
Home-page: https://software.intel.com/en-us/
Author: Intel Corporation
Author-email: scripting@intel.com
License: Proprietary - Intel
Location: /opt/venvs/py37/lib/python3.7/site-packages
Requires: 
Required-by: 

Note that tbb is installed but cannot be imported:

$ python -m tbb --help
/opt/venvs/py37/bin/python: No module named tbb

This is why I forced the use of openmp in the previous example.

Note that I cannot reproduce the problem when running from a conda env with numba and llvmlite installed from the numba channel.