This figure presents the percentage of CI status over the last 90 days. It is limited to master branches, taking into account the route to live only.
As a given pipeline can be triggered several times over a day, only the latest status of the given day is taken into account.
The assumption is that a failing CI pipeline will eventually be fixed if possible at the end of the day. Therefore, the granularity of this figure is one day.
First of all, there is a constant ratio of aborted jobs, which could be hanging jobs being killed by the CI itself, but this does deserve more thorough analysis.
As for the white band on the 26th of August, it corresponds to the absence of data for that day.
Overall, the proportion of failing pipelines is trending upwards and generally speaking at levels which are obviously too high to provide any real confidence in the pipelines themselves.
We need more data to figure out where the jobs are failing. However, the suspicion lies around functional tests. This element has to be confirmed as well.