HT gives you about 20-25% more performance whether it is F9, F10 or SN.
Take my W5580 dual CPU box as an example. Each CPU 4 real cores or 8 HT cores, x 2 CPU, total 16 HT cores.
I usually have 2 nodes running on this box. If WTM is showing less than about 99% usage I might add another node or 2. Just make sure you have enough RAM to run the number of nodes you want to run. An odd number of nodes like 3 nodes for example with 16 cores is fine.
You also don't want too many nodes because if you have a massive scene it gets sent to each node, not each computer, when loading a scene & even with gigabit ethernet it can get painfully slow.
You don't need run more than one node on a render node. I wouldn't recommend running more than one because you can run in to memory problems fast especially when you have a lot of threads/cores . LWSN is multi threaded just like anf9 f10 render.