When you create a node down event, you specify the maximum percentage of nodes that can be down in a clique before the event triggers.
When a node goes down, its vprocs migrate, increasing the amount of work on active nodes. This extra work slows performance. In that degraded situation, you may want to throttle lower-priority requests or enable filters to ensure that critical requests can still meet their SLGs. You may also want to send a notification so that you can take follow-up actions.
Recommendation: Your system may be designed to run with some degradation (a large system with hundreds of nodes may be sized expecting that there is always one node down somewhere). In that case, Teradata recommends that you set the threshold so that a node down event occurs only when the degradation exceeds what the system can absorb.
If your system is not sized to expect down nodes (as is the case with many small-to-medium-sized systems): a good threshold for a down node event is the default, 24%.