Pick one rule and write it down: either a person can only ever be a first MQL, or they can re-MQL after a fixed window (90 days is the common choice). Report net-new and re-engaged separately by default — a single blended MQL number hides which motion is actually working.
The reporting consequence is larger than it looks. If a two-year-old first MQL keeps suppressing new qualification, cohort conversion rates drift downward for reasons that have nothing to do with performance. Decide whether the analysis is cohort-based (first MQL only) or period-based (every qualification event) before you build the report.
Every claim above comes from work LeanScale published. These are the sources.