Those adding "*workercount" and creating seperat "params" in the call fixed the problem:

    params = [(X_train, Y_train, X_test, Y_test, classifier, ...) for classifier, parameters in classifiersAndParameters for mitigator in mitigators for constraint in constraints] * workercount
with multiprocessing.Pool(processes=workercount) as pool:
    params = params[:workercount]
    results = pool.map(workers, params)