It didn't in my tests - also tried sync/copy to/from local.
Incorrect, it always makes a difference whether you need to download 64KB or 32MB (to display a simple icon). 32MB still takes 300ms on a 1Gbps fiber (best case), that is enough to be noticeable when put on top of overhead and latencies.
My test showed no negative effect of --buffer-size=0 on downloads easily surpassing the need for 8K Cinematic UHD. I also guess it may improve responsiveness on video streaming by making the first data earlier available. I may also help a lot when scanning and creating thumbnails, which are probably random reads. I will leave the testing so somebody with a working cinema setup.
My guess is the 16M buffer size was optimized for copy and sync between remotes, where it still makes a lot of sense. Seems like --buffer-size=0 is better suited for mounts on todays pcs (with GUI, gigabit, multicore, SSD etc).
It also has a great effect on large files, when you are doing random reads - apparently without any negative effect on sequential reads. Size really doesn't matter when doing random reads.
I cannot see the need for VFS read ahead on a a typical pc. It can easily move data with the responsiveness needed to efficiently utilize networked storage up to the available bandwidth.
What are the use cases where VFS read ahead makes a difference?