Trying to help, let’s see if I just make things more confusing 
Let’s ignore plotting completely for now… in order to simplify things.
What is being shared above is that running your Chia farm (that includes harvester, farmer, full-node, and wallet) on a machine that talks to NAS storage over the network is likely to have problems. This is because each challenge can require something like 64 disk reads PER PLOT to validate. This “chatter” over TCP/IP to a NAS that may also have power-saving features to spin down idle disks, is what generally leads to slow performance.
As mentioned above, some more power NAS (like some of the Synology units with “+”) can run VMs, so you can offload the harvester process to run locally on the NAS. This then responds to challenges from your farmer process that would run on a dedicated linux machine along with full node, etc. This deployment model would probably work, but I haven’t tried it.
Here is a thread that discusses the differences between Farmers and Harvesters and the relationship between these processes: What's the difference between harvesting and farming? - #3 by roybot