The problem is that there is no one optimum number. This number should depend on what H/W you have (i.e., adjusted during the installation) and what network conditions are at the moment (e.g., reduced during the heavy dust storms (to reduce extra network traffic), or syncing from scratch (as your node is considered dead)).
Instead, someone used his/her grandmother age, and put that value at 80, and no one questions it (as by not questioning it, no one needs to do the work). If the code was properly written, there would be no such number, as program would adjust it on the fly. Because it doesn’t do it, some nodes are falsely claimed to be “low spec” and scapegoated for crap that happens during the dust storm (start_full_node process choking a single core).
So, if you are syncing from scratch, you can put it at 5-10. When you are up and running, try to see the CPU load (per core, not overall), and if low, bump it up (to be a friendly node). In case there is a heavy storm, bring it down to ~10, as from the network point of view it is better to have a node with few connections rather than a dead node.
You can change it in config.yaml
full_node:
...
target_outbound_peer_count: 40