I have noticed that the changes made to this line may introduce unwanted behavior in the future dataframe:

This will make make_future_dataframe() ignore the dates with NaN values in the history, considering

history = df[df['y'].notnull()].copy()

in the versions prior to 1.1.5 (for reference):

self.history_dates = pd.to_datetime(pd.Series(df['ds'].unique(), name='ds')).sort_values()

Notice that the dataframe df['ds'].unique()is used instead of the not null history['ds'].unique()