I agree with the linewidth.
You could also consider that perhaps a clustermap is not the best tool in your case, considering time series. Perhaps (probably) an autocorrelogram would be the better choice.