A bank runs voice over two transports, preferring low-delay MPLS color gold but tolerating broadband color blue, and requires loss under 2%, latency under 150ms, and jitter under 30ms before moving calls. The policy must measure performance and steer per tunnel. Which app-route policy with SLA class criteria and preferred colors meets the voice need?

The correct policy succeeds because voice needs continuous measurement of loss, latency and jitter together, and an app-route SLA class uses BFD telemetry to assess each tunnel independently and steer matching traffic to preferred gold while retaining blue for automatic failover when any bound is breached. A gold-only variant with complete thresholds is legitimate where regulation forbids internet transport, but it fails this scenario conceptually by eliminating required broadband tolerance, so calls would fail during a gold degradation even when blue could sustain them. A loss-and-latency-only variant with correct gold-then-blue order is common for data applications insensitive to interpacket variation, but it fails for voice conceptually because jitter-induced poor audio can remain undetected and calls stay on an impaired tunnel. A full-threshold variant that prefers blue with gold as backup is standard for internet-first cost optimization, but it fails here conceptually by reversing explicit business intent to prefer low-delay MPLS, placing calls on broadband even when gold is healthy. As an exam caveat, presence of colors or numeric limits alone does not imply SLA-driven steering, which only app-route provides, unlike static forwarding in data policy or advertisement filtering in control policy. As a practical check, confirm the SLA values, verify gold is listed before blue in the app-route sequence, then generate impairment and watch tunnel SLA state and voice forwarding shift.