Failure Domain. The bounded part of a system that is allowed to slow down, stop, queue, shed work, or enter degraded mode when a dependency or component fails. A failure domain names the blast radius before the incident does.
Failure domains matter because degradation is only graceful when unrelated work can continue. The boundary should say what depends on the failing part, what remains available, which circuit breakers, queues, kill switches, or fallbacks contain the damage, and which truth and safety guarantees are never allowed to degrade.