The article is mostly concerned with the author not understanding BTRFS and ranting about it. To a degree, that one must assume, the author intentionally gave up early on understanding and simply rants about everything that does not look or work as usual.
The author seems to conflate du and df. They operate completely differently, du calculates the usage of files, df gives a quick estimate of free storage from the file system. Of course df is going to be 7000x (or whatever) faster.
That you need to use du to get per subvolume disk usage is the problem here, I think you missed the point I was trying to me. But that's on me, I couldn't imagine someone would seriously consider that I didn't know the difference between df and du...
Would not read again.