Sometimes, Job failures widget in MR is out of sync
From slack, it seems that when people change tabs and come back, the count of jobs and the widget are out of sync. This is probably because the count comes from REST and the list from GraphQL. Also, the REST call use Visiblity
so I am pretty sure there might be something around a user going to a different tab, coming back and the pipeline is no longer active, so the polling does not resume and data is out of date.
Edited by Frédéric Caplette