Using the script is a lot better bandwidth wise... and it does something interesting that nothing else does currently. It moves all of the packages the core knows about to the /usr/pluto/deb-cache directory, and creates two new Packages.gz files for the two directories. This means when mds are created, they draw all packages from the local directory while being setup. The script also cleans up old versions, so you don't end up with 100 different versions of packages.
So lets say you run the upgrade script on the core and remaining MDs. You could create a new MD after that, diskless setup will go much faster, and then running global upgrades again after creation would bring that MD up to speed.
If you do not care about time and bandwidth, you could upgrade the whole system, and then run Diskless_CreateTBZ.sh. This would have future MDs already up to date.
From my perspective, the result should be the same. One takes about half an hour, one can take up to two hours. One downloads a max of 525Mb, one downloads around 1.2Gb. However, running TBZ means you do not have to run a very immature upgrade script every time you plug an MD in.
So... I guess it comes down to how often you upgrade. If you upgrade a lot... I would use the upgrade script. If you upgrade until you get to a stable image and then hang there... I would run TBZ.