I was testing this out for myself and found that the script doesn't like being pointed at a mergerfs directory, which is likely what some others unknowingly encountered. When done as such it will run through the entirety of the script successfully, however upon checking memory stats before, during and after and then running it again it is clear it's not caching to ram as intended.
It does however work as intended when bypassing mergerfs and pointing directly to the drives mount locations. The thing I'm unsure of however, is whether or not it still gives benefit if Plex/Emby/Jellyfin is trying to read from the mergerfs mount point instead of the direct drive mount locations.