He gave you a really good reason: knocking out most severe, low-level errors in high-usage, critical services. You'll put work in for the risky ones. Why not the low-risk alternatives...