Which mechanism provides fast path failure detection?

Non-Stop Forwarding maintains the forwarding plane during a control-plane restart or failover, but it masks failures rather than detecting them.

Carrier delay introduces a configurable delay before the routing protocol reacts to an interface state change, which slows down - not speeds up - failure detection.

Graceful restart allows a routing protocol to restart without tearing down adjacencies to preserve forwarding continuity; it is a resiliency mechanism, not a failure detection mechanism.

UDLD detects unidirectional fiber or copper link conditions at Layer 2 but operates on longer timers and is scoped to physical link issues, not general routing path failures.

Fast hello packets configure sub-second hello timers on routing protocols such as OSPF and IS-IS, allowing the network to detect adjacency loss in milliseconds rather than waiting for the default dead interval (typically 40 seconds). This directly provides fast path failure detection by accelerating the neighbor keepalive mechanism. Among the listed choices, fast hello packets are the explicit mechanism purpose-built for rapid failure detection.

iSPF speeds up SPF recalculation after a topology change is already known but does not itself detect that a failure has occurred.