netCDF4-python dataset objects getting garbaged collected inside context manager?

test_open.py

#!/usr/bin/env conda-execute
# coding=utf-8
"""
Testing netCDF4 garbage collection differences between 2.7 and 3.4

**Fails**
    - python 3.4 / netCDF4-python 1.2.1, 1.2.0, 1.1.9, and 1.1.8

**Passes**
    - python 3.4 / netCDF4-python 1.1.6
    - python 2.7 / netCDF4-python 1.2.1, 1.2.0, 1.1.9, 1.1.8, 1.1.6
"""

# conda execute
# env:
#   - python 3.4.*
#   - netcdf4 1.2.1
#   - cf_units 1.0.0
# channels:
#   - ioos
# run_with: python

import netCDF4
import cf_units as unit

units = [ unit.Unit(x) for x in [
    '%',
    'A',
    '1e-3',
    'pascals',
    'V',
    'W.m-2',
    'cm.s-1',
    'degrees',
    '1',
    'ft',
    'ft.s-1',
    'ft3.s-1',
    'hectopascal',
    'inch_Hg',
    'inch',
    'kg.m-2',
    'kg.m-3',
    'knot',
    'm',
    'm.s-1',
    'm3.s-1',
    'mS.cm-1',
    'millibars',
    'mg.L-1',
    'mm',
    'mmole.m-2',
    'mile.hour-1',
    'pH',
    '.001',
    '1e-6',
    's',
    'degree_Celsius',
    'degree_Fahrenheit',
    'microatm',
    'microg.L-1'
]]


with netCDF4.Dataset('axiom_cruise_profiles.nc', 'a') as nc:
    for vname in nc.variables.keys():
        v = nc.variables[vname]
        if not hasattr(v, 'units'):
            continue
        try:
            un = unit.Unit(v.units)
        except ValueError:
            continue
        for oku in units:
            if un.is_convertible(oku):
                v[:] = un.convert(v[:], oku)
                print("Converted {} from {} to {}".format(v.name, v.units, oku.format(unit.UT_NAMES)))
                v.units = oku.format(unit.UT_NAMES)
                break