if you use rclone mount, then you do use VFS. not sure what you are talking about.
vfs cache modes can be enabled for when you need to read/write to a file at the same time. I have encountered some software that needs this (bazarr is an example).
Here an example of my rclone mount command for my encrypted volume:
rclone mount gcrypt: /mnt/gcrypt --uid=1000 --gid=1000 --umask=002 --timeout=1h --tpslimit=10 --user-agent=RandomAgent --dir-cache-time=1440m --vfs-cache-mode=writes --vfs-cache-max-age=72h --vfs-cache-max-size=200G --vfs-read-chunk-size-limit=off --vfs-read-chunk-size=128M --rc --buffer-size=1024M
I also pair this mount with mergerfs using a command like the following:
mergerfs -o cache.files=auto-full,dropcacheonclose=true,use_ino,func.getattr=newest,category.create=ff,minfreespace=0,fsname=unionfs /mnt/move=RW:/mnt/gdrive=NC:/mnt/gcrypt=NC /mnt/gmedia
This works very well for me. I have a very large system - 120k TV episodes (215TB), 15k movies (200TB). My server is pretty decent (e3-1285v6, 64GB RAM, 4TB NVME, 1Gbps internet)
It is easy to hit API limits with my setup since it is so large. I use sonarr/radarr and there are settings in there that will analyze media (I turn that off). Bazarr grabs subtitles for me, it has a feature to scan in the media for subtitles and I had that turned off until recently. Generating thumbnails, performing extensive media analysis, scanning library automatically, generate intro video markers are all features in plex that can quickly have you hit your API max.
I am really looking forward to the new VFS caching features that are currently being beta tested. I feel that this might be the rclone feature that allows me to turn on some of these additional features mentioned above without hitting API limits.
also, my start time is nearly instantaneous, less than 1sec. When they aren't instantaneous, I know that I am having an API issue and that something has changed. Recently that change was the detection of tv intros, which I quickly disabled.