Currently semversioner check asserts that a change document has been created. And initially it works great for that, but imagine this flow:

In PR 1, Cyril makes low-impact changes. semversioner check catches that he needs a change document. Cyril uses semversioner add-change -t patch -d ...

In PR2, Pat makes a higher-impact change. They forget to commit a change document, but semversioner check will pass, because a change document exists from Cyril's PR.

Ideally, semversioner check would cause Pat's PR to fail as well because their changes have semver impact that should be documented independently.