How do sklearn SVC hyperparameters affect training time?

I'm using skopt BayesSearchCV, although I previously had the same issue using Keras-Tuner Bayesian Oracle. I'm trying to optimize the hyperparameters of an SVC. There are 1200 data points in the train dataset with only 5 features each, so dataset size shouldn't be the issue. Some of the models train in a fraction of a second while some just never finish training, so I assume the bounds for my hyperparameters need to be adjusted. Currently I have:

model = pipeline.Pipeline([('scaler', StandardScaler()), ('model', SVC(cache_size=500))])

search_params = {
    "model__C": Real(1e-4, 1e4, prior='log-uniform'),
    "model__kernel": Categorical(['linear', 'poly', 'rbf', 'sigmoid']),
    'model__degree': Integer(2, 8),
    'model__gamma': Real(0.01, 0.9),
    'model__coef0': Real(0.001, 1000, prior='log-uniform'),
    'model__class_weight': Categorical([None, 'balanced'])
}

searchcv = BayesSearchCV(
    estimator=model,
    search_spaces=search_params,
    n_iter=100,
    n_jobs=-1,
    cv=5,
    refit=True,
    error_score=0,
    scoring=metrics.make_scorer(metrics.balanced_accuracy_score),
    return_train_score=True,
    verbose=1
)
searchcv.fit(x_train, y_train) 

How should I change my hyperparameter spaces so that it doesn't get stuck training a single model for hours