Python Parallelize Complex Multi-object for

Question

I've got a multi-object loop like this:

a1 = A('path1')  # Class from external library
a2 = A('path2')  # Class from external library
b = B('path3', a2)  # Class from external library

for first, second in foo(a2, a1):  # function from external library
    # this loop is executed more than 100.000 times      
    # and has code that implies multiple conditions
    # writing in files and so on

I tried with joblib, ray and dask but unsuccessfully (serialization errors and so on).

What approach would you use? Does it make sense?

Thank you in advance for your help