Our first attempt at follow-the-sun on-call just moved the burnout around the globe instead of eliminating it. Here's the version that finally stuck.

Follow-the-sun sounds like a solved problem: three regions, eight-hour windows, nobody ever pages at 3am local time. Our first attempt technically achieved that and still produced complaints nearly identical to the single-region rotation it replaced.

Why the first version failed

We'd solved for time-of-day but not for handoff quality. Each region was working from the same shared incident queue with minimal structured handoff, so engineers in the next region routinely spent the first hour of their shift re-diagnosing something the previous region already understood, just because the context lived in someone's head instead of on paper. The pager stopped ringing at 3am. The job got harder in a different way instead.

What the working version does differently

We made handoff a blocking step, not a courtesy — the outgoing region cannot formally end its shift in the tool until a structured handoff is filed, and the incoming region cannot start without acknowledging it.

shift_transition_policy: follow_the_sun_v2
regions: [AUS(UTC+10), EU(UTC+1), US(UTC-6)]
window_hours: 8, with 30-min overlap for live handoff call if any P1/P2 open
handoff_required_fields:
  - open_investigations: [ticket_id, current_hypothesis, next_step]
  - watch_list: [metric, threshold, why_watching]
  - explicitly_resolved_since_last_handoff: [ticket_id, resolution]
enforcement: >
  Outgoing engineer cannot close their shift in the on-call tool
  without all fields populated. Empty fields block, not warn.
"The 30-minute overlap window felt like overhead until the first time it saved someone forty minutes of rediscovering what I already knew."

What changed once handoff was enforced

Time-to-first-productive-action after a shift change dropped by roughly 35%, and — more relevant to culture — engagement survey comments about follow-the-sun on-call shifted from "exhausting" and "confusing" to neutral-to-positive within two quarters. Removing the 3am pages was necessary but not sufficient. Removing the redundant re-diagnosis was what actually made the rotation feel sustainable.

We also learned to resist skipping the overlap window when a queue looked quiet. Our two worst handoff failures happened precisely because a region judged things calm enough to skip the live call, only for a slow-building issue to surface an hour into the next region's shift with none of the earlier context written down anywhere.

  • Follow-the-sun eliminates bad hours but can silently replace them with redundant re-investigation work if handoffs aren't structured.
  • Make handoff a blocking gate in your tooling, not an optional courtesy — courtesy decays under pressure.
  • A short live overlap window for open P1/P2s is worth the calendar cost; async notes alone don't capture urgency well.