Reduce apdex threshold for CI runners

Context: gitlab-com/gl-infra/production#3800 (closed), gitlab-org&5434 (closed)

We are currently experiencing pressure on the CI runners service in a few ways. A rapid action is underway to improve performance, and we are undertaking some mitigations in the related incident issue. Retreating from 97% to 95% allows us to spot further regressions, but the current pager load is broadly unactionable (in a short-term sense).

Merge request reports

Loading