I am not an NFS expert, but I think kernel threads are different from the nfsd (which is a userspace process which in turn hands over requests to kernel). In any case, I believe the nfsd threads you referred may be more relevant in the general purpose operating system, where you tend to map a thread with number of CPU cores. However, ONTAP (CSMP) is a optimized storage operating system, all the processes runs as per their domain distribution in a highly optimized and parallel manner. I would suggest reading the following TR, it covers lot of useful information related to performance tuning for NFS specifically.
Have a look here:
Page 96/97:Nconnect
Page 120:Exec context throttling
Page 122:Network connection concurrency and TCP slots
https://www.netapp.com/media/10720-tr-4067.pdf
Quoting from the pdf: The purpose of nconnect is to provide multiple transport connections per TCP connection or mount point on a client. This helps increase parallelism and performance for NFS mounts.
CSMP:
https://kb.netapp.com/Advice_and_Troubleshooting/Data_Storage_Software/ONTAP_OS/CPU_as_a_compute_resource