This incident has been resolved.

This was an unusually complex issue, one we haven’t encountered before. The symptoms appeared only for teams in UTC+1 and all time zones westward (down to UTC-13), while everything from UTC+2 through UTC+14 worked perfectly. This made it particularly tricky to reproduce and diagnose.

The fix itself was straightforward, but the investigation took time. We added additional checks when calculating slot boundaries in weekly rotations. We also corrected how offsets were carried forward after several weeks, which was the main factor causing schedules to drift.

You should now see accurate on-call assignments across all time zones. We’ll continue observing these schedules over the next few days.

We believe we have identified the root cause of the on-call display issue and, unsurprisingly, it’s related to timezones.

Our recent changes to make rotation-day adjustments easier introduced a small weekly shift for schedules with weekly rotations. Over time, this shifted enough to create a full-day gap in certain on-call layers.

We are currently testing a patch and are already seeing improvements in our internal checks. More updates to come soon.

We are currently investigating an issue where the on-call schedule is not showing the correct on-call person in the Spike UI as well as in exported Google Calendar events.

Our team is also reviewing related API responses to determine if there are any inconsistencies there.