We explicitly set the end_date of VWAPBestEffort orders to the close.

The internal code that handles the start/end dates is roughly as follows:

    def _adjust_date_bounds(start_date, end_date):  
        """  
        Helper method for adjusting start and end time bounds.  
        """  
        current_time = get_datetime()  
        _, todays_close = (  
            trading_environment.get_open_and_close(current_time)  
        )

        if current_time >= todays_close - timedelta(minutes=1):  
            raise UnsupportedOrderParameters(  
                msg="Can't place a VWAP order in the last minute of the day."  
            )

        # Validate start date  
        if start_date and start_date < current_time + timedelta(minutes=1):  
            raise UnsupportedOrderParameters(  
                msg="start_date must be at least one minute "  
                "later than the current algo time."  
            )  
        start_date = start_date or (current_time + timedelta(minutes=1))

        # Validate end date  
        if end_date and end_date > todays_close:  
            raise UnsupportedOrderParameters(  
                msg="end_date must be <= market close for current day"  
            )  
        end_date = end_date or todays_close

        if start_date >= end_date:  
            raise UnsupportedOrderParameters(  
                msg="start_date must be before end_date"  
            )

        return start_date, end_date  

But I thought orders are sent to IB asysynchronously/immediately as they are encountered in the code. Is VWAPBestEffort different?

Orders are sent asynchronously, but as you can see in the above, we shift the start_date of the order forward to the next minute. I think this is because IB will reject the VWAP order if the start date is before the current real time, but I'm not 100% sure.

Does IB support VWAPBestEffort for their paper trading?

VWAP orders won't be rejected by IB, but I don't know how well they're able to simulate actual behavior or how hard they try to do so.

Has IB published the details of their execution algorithm?

The general gist seems to be to buy a percentage of the full order on each bar between the start and end dates of the order, but I don't believe the're a formal publication of the details anywhere.

Any plans to emulate IB VWAPBestEffort for backtesting and paper trading?

Probably someday, but also probably not soon. The machinery to make this work lives in Zipline though, so contributions welcome!