Resilience against weird failures and denial of service attacks

In networking theory a network connection could be successful as long as there is at least one operational path between the source and the target. It should not matter how many failed network devices and lines there are and in which way they have failed. This includes failures which have been deliberately caused by a human to harm the network.
However there are some failures which are harder to recover from:

Layer 2

If multiple hosts are connected to the same network segment, it is usually easy for a single host to dos the segment. There is little cor can do to change this. Switched networks can be partially protected with some hardware support by manually assigning mac addresses to switch ports and filtering invalid mac addresses and excessive broadcast traffic.

However in ad-hoc wlan networks this remains unresolved and so an attacker can dos the network locally.