The handoff window nobody can agree on

Follow-the-sun on-call is a good idea. The pitch is obvious: nobody gets paged at 3am, because at any hour there is someone awake and holding the pager.

It works right up until you have to pick the moment the pager changes hands.

Why the handoff is the hard part

The rotation itself is easy — two people, twelve hours each. The handoff is where it breaks, because a clean pass needs both engineers awake and working at the same time, and the whole design premise was that their working hours barely overlap.

Push it to the outgoing engineer's end of day and it lands at the incoming one's breakfast. Push it the other way and someone is writing up an incident at 10pm.

Pick the boundary, then defend it

Three things make the difference between a handoff that holds and one that quietly degrades into "the pager just moves at midnight and nobody talks":

Put it inside the real overlap, not at its edge. If your two zones share four hours, the handoff belongs somewhere in the middle of that, not at minute one or minute 240. Edges get eaten by meetings that run over and mornings that start late.

Make it a fixed local time for one side, and accept the drift on the other. Trying to keep it "fair" for both by moving it around produces a rotation nobody can remember. Pick the side where a fixed time matters most — usually whoever has the denser meeting load — and let the other absorb the seasonal shift.

Give it fifteen real minutes. A handoff is a conversation: what is still open, what was noisy overnight, what is likely to page next. If it is not on the calendar it will not happen, and the incoming engineer will rediscover the overnight incident from scratch.

The transitions will break it

This is the part that bites on-call specifically, because the failure is silent.

If your two engineers are in regions that change clocks on different weekends — the US and the EU do — then twice a year your handoff time shifts by an hour for one of them. For a couple of weeks, the pass either overlaps by an hour more than intended or, worse, gaps by an hour with nobody holding the pager.

If one side does not observe daylight saving at all — India, Arizona, most of Asia — the gap moves without that side doing anything.

Two defences:

• Define the rotation in UTC in whatever tool schedules it. UTC never changes, so the actual coverage boundary stays put and only the humans' local-time experience shifts.
• Diary a check for mid-March and late October. Those are the only two moments this can silently move. See Eastern Time to GMT.

Check the coverage, not just the handoff

The other thing worth verifying once: that your two twelve-hour shifts genuinely cover twenty-four hours in both halves of the year. A rotation designed in January against a five-hour gap has an hour-shaped hole in it during the weeks the gap is four.

Related: We hired someone eight zones away · UTC to EST

Figures on this page are computed from the IANA time zone database. Last generated 2026-08-18.