A customer wants to combine every optimization suggestion from a code review into a single unreviewed rewrite to save time. Why is this the wrong approach?

The correct answer is the one that says combining every optimization into a single unreviewed rewrite removes the ability to isolate which change caused a regression and skips the review that catches problems before they reach production. When several optimizations land together with no individual review, a regression anywhere in the batch cannot be traced to a specific change, and the review step that would have caught it never happened. The option claiming a single rewrite is fine as long as the final build compiles without errors is incorrect because compiling successfully says nothing about whether behaviour changed; regressions in logic or performance are not compile errors. The option saying it is fine provided the rewrite is deployed to a test region first and monitored for a few hours is incorrect because a few hours of monitoring in a test region does not substitute for reviewing each change against its own regression tests. The option saying combining changes is only risky when the application is customer-facing is incorrect because the risk of unreviewed combined changes is the same regardless of whether the application is customer-facing; it is a property of the process, not the audience. For the exam, keep optimization changes small and reviewed individually, never batched into one unreviewed rewrite.