Pauli Decomposition takes a very long time for circuits with nqubit>=8

[thegiantspaceman]

For nqubit=4 the script works well. To confirm this, set nqubit=4 and run "VQLS_github.py"

For nqubit=8 the script gets hung on line 173 at "PauliExpectation(group_paulis=False).convert(measurable_expression)"

Digging into the "PauliExpectation" source code, I find that it is getting hung on the line "pauli_obsv = operator.primitive.to_pauli_op(massive=False)" where it is converting the operator to Pauli matrices.

My questions are,

1. Why does it take so long to convert to Pauli matrices for 8 qubits? An 8 qubit dimension matrix has size 256x256 which is not very large.
2. How can I speed this up so that it is practical to run? Note that my circuit batch size is 32 for 8 qubits.