After a canary showed elevated tool-error rates, you need an immediate switch back to the last known-good multi-agent graph and model deployment with minimal DNS-level complexity. Environments already run side by side. Which approach is most appropriate?
Select an answer to reveal the explanation.
Short Explanation
Answer B. Multi-environment strategies include blue/green and explicit rollback. With side-by-side environments, switch traffic to the last good multi-agent graph/model quickly. Hoping, destroying both stacks, or telling users to wait are not controlled rollback strategies.
Full Explanation
Correct answer: B. Design rollback strategies as part of release management. Blue/green (or deployment slots/traffic manager patterns) lets you flip from the failing agent stack to the prior healthy stack with low RTO. Keep infrastructure, prompts, and evaluation baselines versioned so the green/blue target is truly known-good.
A is incorrect: active tool-error elevation needs decisive rollback, not passive waiting.
C is incorrect: destroying both environments maximizes downtime and recovery uncertainty.
D is incorrect: business workarounds do not restore service and skip engineering accountability.
After rollback, run a defect analysis using correlated traces from the failed canary.