SymptomA scheduled routine returns an empty result every run, phrased as a normal quiet day, and can run unnoticed for weeks.
CauseAn empty result is treated as a successful run. Two distinct triggers are documented: a site layout change that breaks the captured steps, and exhausted third-party credits (for example X) that make the source unreadable.
DetectionHave the routine report its inputs, not only its findings, so a run that read nothing is visibly different from a run that found nothing. Track source freshness explicitly.
FixGive the empty result a named outcome. Add an if nothing, send nothing rule so the Bot cannot invent a justification. Re-test after any website, connector or source-format change, because a green run last month is weak evidence once the source has changed.