Rclone cache speed (test with Disk Speed Test by Blackmagic)

Looking at your log all the io is still being done in 4k chunks.

Can you try some different power of 2 sizes for iosize - ie some of these?

131072
262144
524288
1048576
2097152
4194304
8388608
16777216
33554432

Try 16777216, 8388608 and 131072

You could also try experimenting with this value (without iosize)

  --max-read-ahead SizeSuffix              The number of bytes that can be prefetched for sequential reads. Not supported on Windows. (default 128k)

If you look through the log you'll see

$ grep 'Read: n' rclone155beta-local-iosize-only.log
2021/03/24 11:21:40 DEBUG : /tempAAA/DiskSpeedTestTemp: >Read: n=4096
2021/03/24 11:21:40 DEBUG : /tempAAA/DiskSpeedTestTemp: >Read: n=4096
...

You'll see all the Reads are 4096 - you want them to be bigger!

Likewise the Writes

$ grep 'Write: n' rclone155beta-local-iosize-only.log
2021/03/24 11:21:34 DEBUG : /tempAAA/DiskSpeedTestTemp: >Write: n=4096
2021/03/24 11:21:34 DEBUG : /tempAAA/DiskSpeedTestTemp: >Write: n=4096
...