On-Call Burnout: Warning Signs and How to Fix Your Rotation

On-call burnout doesn’t usually show up as a dramatic resignation letter. It shows up as a slightly worse mood on Mondays, a growing reluctance to volunteer for anything extra, and eventually, a good engineer quietly updating their resume. By the time it’s obvious, you’ve usually already lost something — trust, goodwill, or the person entirely.

Treating on-call burnout as a reliability risk, not just a wellbeing nicety, tends to get it taken more seriously. A team that burns out its on-call rotation doesn’t just lose people — it loses the institutional knowledge that made incidents resolve quickly in the first place.

Warning signs worth watching for

Burnout rarely announces itself. These are the signals that tend to show up first:

• Rising after-hours alert volume with no corresponding rise in real incidents. If pages are increasing but outages aren’t, your alerting is getting noisier, not your systems less stable — and someone is absorbing that noise.
• The same two or three people handling most escalations. Even with a nominally fair rotation, informal patterns emerge where certain people get pulled in disproportionately because they’re fast, senior, or just don’t say no.
• Increasing swap requests or no-shows. Occasional shift swaps are healthy. A steady increase in swap requests, or someone missing a shift they were scheduled for, is usually a sign the rotation has stopped feeling sustainable to them.
• Slower acknowledgment times. Someone who used to respond in two minutes now takes fifteen. That gap is often not carelessness — it’s a person who has started to dread the notification sound.
• Attrition concentrated on the on-call roster. If the people leaving your team disproportionately include your most active on-call responders, that’s not a coincidence worth ignoring.

Why it happens

Burnout is rarely caused by one bad night — it’s the accumulation of several structural problems:

Uneven distribution. Rotations that look fair on paper (everyone gets one week a month) can still be unfair in practice if some services page far more than others and the same person always ends up covering the noisiest ones.

Low signal-to-noise ratio in alerts. Every alert that didn’t need a human response trains the on-call engineer to distrust the next one — and to feel the weight of every page a little more, since they can no longer assume it matters.

Unclear escalation ownership. When it’s ambiguous who’s supposed to respond, or the escalation policy doesn’t actually work, the person on-call ends up absorbing responsibility beyond what their shift was supposed to cover.

No recovery time after a bad night. Getting paged three times overnight and then being expected at a 9am standup, repeatedly, adds up in a way that a single rough night never does.

Fixes that actually help

Rotate fairly, and check the math, not just the calendar. A rotation can be evenly scheduled by week and still be unevenly distributed by pain. Periodically audit who’s actually absorbing the volume, not just whose name is on the calendar.

Tune alerts down to what’s actionable. If an alert fires and nobody ever does anything in response, it shouldn’t be paging a human — route it to a dashboard or a lower-urgency channel instead. Cutting noise is often the single highest-leverage fix available.

Set and enforce quiet hours or a rest period. If someone gets paged overnight, build in an expectation that they get a slower morning or an afternoon off, rather than showing up to standup running on three hours of sleep like nothing happened.

Track load per person, not just per team. A quarterly look at who got paged, how often, and at what hours turns a vague sense that “on-call is rough right now” into something you can actually act on.

Consider compensating on-call time. Whether it’s stipend pay, time off in lieu, or simply lighter sprint commitments during on-call weeks, acknowledging that on-call is real work — even the weeks nothing goes wrong — goes a long way toward making it feel sustainable rather than extracted.

None of these fixes require a complicated program. Most of them just require someone to actually look at the data behind the rotation instead of assuming it’s fine because nobody’s complained loudly yet. The tooling helps too — fair rotation scheduling, escalation policies that actually work, and alerts that reach the right channel on the first try all reduce the day-to-day weight of being on-call. But the biggest lever is usually just paying attention before the warning signs turn into an exit interview.