Which team-level delivery metrics have changed a small team's behaviour for the better, and how were they retired?

Open question

Recommendations for team metrics usually come from large organisations with dedicated tooling. DORA's guide defines five software delivery metrics (change lead time, deployment frequency, failed deployment recovery time, change fail rate and deployment rework rate) and notes that they are best applied to one application or service at a time and in context. A small team can compute these, plus carried-over items, review wait time and pages per shift, from its own trackers and pipeline, but each number costs attention to collect and to discuss, and any number that becomes a target changes the behaviour it was meant to observe. What is missing is documented experience from small teams: which single metric, if any, led to a concrete change (a process change, a staffing decision, a stopped project), how long the metric stayed useful, what the first sign of gaming looked like, and what the team did to retire it without losing the habit of measuring anything.

What a useful answer contains

Team size and domain; the metric's exact definition and how it was collected (manually, from the tracker, from the deploy pipeline); the decision or change it triggered, with dates; whether the team looked at trends or thresholds; observed gaming or distortion, if any, and how it was noticed; how and why the metric was dropped or replaced; and whether the practice survived a change of lead or of team composition. Reports of metrics that were collected and never acted on are as useful as success stories, and should be labelled as such.