Bug: Refactorization fails when used after 0th matrix in the ACTIVs series

[shakedregev]

Description how to reproduce the bug

This reproduces the error (making refactorization happen after 0-th matrix, instead of 1-st)

./examples/kluRefactor.exe -i -m ~/ReSolve_dir/testcases/ACTIVSg10k_AC/matrix_ACTIVSg10k_AC_ -r ~/ReSolve_dir/testcases/ACTIVSg10k_AC/rhs_ACTIVSg10k_AC_ -n 3 
System 0:

========================================================================================================================
Reading: /home/szb/ReSolve_dir/testcases/ACTIVSg10k_AC/matrix_ACTIVSg10k_AC_00.mtx
========================================================================================================================

Finished reading the matrix and rhs, size: 238072 x 238072, nnz: 1209368, symmetric? 1, Expanded? 1
COO to CSR completed. Expanded NNZ: 1209368
KLU analysis status: 0
KLU factorization status: 0
KLU solve status: 0
	2-Norm of the residual: 1.0053899214548057e-11
FGMRES: init nrm: 1.0053899214548057e-11 final nrm: 6.9530826398253198e-12 iter: 1
System 1:

========================================================================================================================
Reading: /home/szb/ReSolve_dir/testcases/ACTIVSg10k_AC/matrix_ACTIVSg10k_AC_01.mtx
========================================================================================================================

Finished reading the matrix and rhs, size: 238072 x 238072, nnz: 1209368, symmetric? 1, Expanded? 1
COO to CSR completed. Expanded NNZ: 1209368
KLU re-factorization status: 0
KLU solve status: 0
	2-Norm of the residual: -nan
System 2:

========================================================================================================================
Reading: /home/szb/ReSolve_dir/testcases/ACTIVSg10k_AC/matrix_ACTIVSg10k_AC_02.mtx
========================================================================================================================

Finished reading the matrix and rhs, size: 238072 x 238072, nnz: 1209368, symmetric? 1, Expanded? 1
COO to CSR completed. Expanded NNZ: 1209368
KLU re-factorization status: 0
KLU solve status: 0
	2-Norm of the residual: -nan

Re::Solve version

All versions have the problem, but we worked around it by refactorizing only starting from 2nd matrix

System and environment details

Fails on all backends

Additional information

No response

[shakedregev]

The 0-th matrix has a much smaller condition number than other matrices: $O(10^7)$ instead of $\geq O(10^9)$
This means it might not be a bug, just a difficulty of the matrix series.
As far as I can tell, GLU only gets around this bug by doing a full factorization under the hood.