Why does scipy.sparse.linalg.spsolve crash when numpy.linalg.solve does not?

Question

I have a sparse Matrix B and an array b (uploaded here), and am trying to do

import numpy as np
import config
from scipy.sparse.linalg import spsolve
from scipy.sparse import load_npz
from numpy.linalg import solve

B = load_npz('B.npz')
b = np.load('b.npy')

spsolve(B, b)

which leads to

corrupted size vs. prev_size
Process finished with exit code 134 (interrupted by signal 6:SIGABRT)

the same problem does not happen when I use numpy.linalg.solve:

B_dense = np.array(B.todense())
solve(B_dense, b)
Out[5]: 
array([-299.77073372, -299.68350884, -299.59972996, ..., -286.52926438,
   -286.50847706, -286.48927409])

So since numpy's version works, I suppose nothing is "wrong" with my matrices? Why does this occur? And since this crashes the code without a trace -- is there perhaps a way to "wrap" the error in an exception, to run solve whenever spsolve would lead to an issue?