I think I understand your set up more or less ... i don't quite get the part you describe how your process its sped up by enabling cache; "moving from one sector to another on the SSD."
They way I do it, I keep all of the, lets call it "workload" on the cache drive, all the time. Cache witch is typically a SSD device. Your process has the workload part on the cache, part on the array because each time the mover moves, it moves some of your unprocessed rar files to the slower spinners.
Next phase when Sabnzb (or whatever other docker you use for that matter) starts with unpacking its not on your super fast SSD but on the slower spinners. To make matters worsen your CPU needs to calculate parity witch causes even more load you can easily avoid.
Final point is that unRAID really excels when you can organize its writes to the array sequential. In other word one by one. My workflow does that.
I'm not saying yours is no good, I'm only offering a way to optimize
edit: when reading it back, i come to the conclusion bjp999 said the same as me