multiprocessing Pool().apply_async(Portfolio.from_signals())

[MaxRepin]

Hi all
I've tried to use multiprocessing with Portfolio.from_signals() for different chunks of exit signals. Thought this makes sense as the chunks don't depend on one another.
However the results of total_return() are much larger than when I run it in a single process (e.g. 200%+ vs 120%), although if I sort the results their order is the same as in the single process - so it looks like some extra fixed 'addition' is made somewhere.

total_returns, chunk_returns = [], []
with Pool() as p:
        for stop in vals:
            mask = sl_stops.columns.get_level_values('stop_value') == stop
            chunk_stops = sl_stops.loc[:, mask]
            args = ohlcv['Open'], entries, chunk_stops
            chunk_returns += [p.apply_async(get_total_return, args)]
            if len(chunk_returns) == 4 or (stop == vals[-1] and chunk_returns):
                total_returns += [r.get() for r in chunk_returns]
                chunk_returns = []

def get_total_return(close, short_entries, short_exits):
        dema_pf = vbt.Portfolio.from_signals(close=close,
                                             short_entries=short_entries,
                                             short_exits=short_exits,
                                             freq='1d')
        tr = dema_pf.total_return()
        del dema_pf
        gc.collect()
        return tr

I know that multiprocessing support isn't declared, just thought it'd be useful and save time for someone else.