wrf.interplevel returns NaN when interpolating at a level where the sample points are

Hi, I recently used wrf.interplevel to interpolate a 3d field onto a single pressure level 50000 Pa. I found a very wired result that every time when the 50000 Pa appears in the sample points, the result is NaN. If I added a little value to 50000Pa, it returns the correct value (approximately). Below is just a simple example to reproduce the issue. Please help!

# construct vert array
xx = np.linspace(0,1,11)
xg = np.zeros((11,4,3))
for j in np.arange(4):
for i in np.arange(3):
xg[:,j,i] = xx

# assign field3d array
yg = np.random.rand(11,4,3)

# print results for comparison
print(' True answer: ', yg[5,:,:])
print(' NaN result: ', wrf.interplevel(yg,xg,0.5).to_numpy())
print('Corrected result: ', wrf.interplevel(yg,xg,0.5+1e-8).to_numpy())

Start by running the provided NumPy and wrf.interplevel example, then trace the interplevel implementation and its handling of an exact sample level. Compare the result at 0.5 with yg[5,:,:] and with the 0.5+1e-8 case. Done means exact-level interpolation no longer produces NaN and the existing behavior remains correct for nearby levels.