The process was decided to be stacks of stacks early on, For the first test object of eagle nebula about 1 year back I got about 17,000 non lucky subs from people to answer if we needed every sub, or only well done stacks
Once you have that stack it gets combined with others with some very detailed pixinsight tricks to keep it in linear space as much as possible, the write up for the image on astrobin goes into the process used for that image.
How it started was I took a question too seriously and kick started the BAT almost 1 year ago, working out what the limits where and what programs could be used, but now its evolved into its own thing with smarter people than me
the paper for resolution improvement vs time has some extra factors relating to apeture, but for the 6-14 inch for amateur's its close enough to not worry about,
He did actually show the stacked lucky image in the video when he was talking at the end on discord, that and some other lucky contributions were given a higher weight than the deeper data from what I can understand to preserve there details,