contour not closed across 0 meridian in polar projection

Hi all,
I'm trying to fetch contour line coordinates from a npaeqd projection plot and I noticed that the contour lines will be broken into 2 parts whenever it crosses the 0-degree longitude, even though they form a closed contour and after calling addcyclic(). Below is minimal working example:

import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.basemap import addcyclic
from mpl_toolkits.basemap import Basemap

lats=np.linspace(0,90,90)
lons=np.linspace(0,360,360)

# make some toy data
xx,yy=np.meshgrid(lons,lats)
z=np.cos(xx*np.pi/180)*np.sin(yy*np.pi/180)

# add cyclic
z,lons=addcyclic(z,lons)
xx,yy=np.meshgrid(lons,lats)

# get contours
bmap=Basemap(projection='npaeqd',boundinglat=0,lon_0=0,
        resolution='l')

fig=plt.figure(figsize=(12,6))
ax1=fig.add_subplot(1,2,1)

contours=bmap.contour(xx,yy,z,[-0.6,0.6],latlon=True,ax=ax1)
bmap.drawcoastlines(ax=ax1)

clines1=contours.collections[0].get_paths()
clines2=contours.collections[1].get_paths()

print 'len(clines1), num of contours across 180', len(clines1)
print 'len(clines2), num of contours across 0', len(clines2)

# plot contours
ax2=fig.add_subplot(1,2,2)

xs=clines1[0].vertices[:,0]
ys=clines1[0].vertices[:,1]
ax2.plot(xs,ys,'b-',label='Contour across 180')

xs=clines2[0].vertices[:,0]
ys=clines2[0].vertices[:,1]
ax2.plot(xs,ys,'r-',label='Half contour across 0')

xs=clines2[1].vertices[:,0]
ys=clines2[1].vertices[:,1]
ax2.plot(xs,ys,'g-',label='Half contour across 0')

ax2.legend()

plt.show(block=False)

Figure output here
The yellow contour on the left are made up by 2 lines (red+green) on the right. This makes it difficult when I try to detect and track some features that move across the 0-meridian.
Is it intended or a bug?

Some specs:
basemap 1.0.7
matplotlib 2.2.2, both installed via conda install

Start with the minimal example using addcyclic() and Basemap.contour() in the npaeqd projection, then inspect how contour paths are returned around the 0-degree meridian. Done means the closed contour is returned as one continuous path rather than two parts, while the example's contour across 180 degrees remains correct.