Burnout Is a Systems Failure

Thesis

When output collapses after sustained over-utilization, the correct postmortem target is the work system: intake, pacing, recognition, safety. Resilience training is what organizations buy when they would rather not change the graph. The person is the symptom. The environment is the incident.

Context

I have watched this plot across three organizations in the last decade. A high-output contributor inherits work that scales faster than headcount. They ship. Intake increases. Somewhere between months nine and fourteen, throughput drops, cynicism rises, and they leave or take medical leave. Leadership frames it as disengagement. The utilization graph, if anyone pulled it, would show a resource run hot until failure — the same curve as an oversubscribed database.¹

Headcount frozen. Roadmap unchanged. The third engineer in eighteen months exits after a “performance” conversation. Names stay out of the writeup because the writeup is about the rotation, not the people who paid for it. Personal failure is cheaper to administrate than structural change. That accounting trick is how the incident keeps recurring.

Christina Maslach’s research treats burnout as an occupational phenomenon: exhaustion, cynicism, reduced efficacy, driven by workload, control, reward, community, fairness, and values. When those variables skew wrong for long enough, people break in predictable ways. Character is not the independent variable.²

Mechanism

Wellness programs are downstream patches. Meditation stipends and resilience workshops are not harmful. They are often substitutes for upstream work: reducing work in progress, renegotiating dates, staffing to a pace a human can sustain for a year. Organizations buy sandbags because moving the river is politically expensive. The sandbags photograph well. The river still floods.

In ops, a service that fails three times in a quarter triggers a blameless postmortem. In people systems, three departures in eighteen months trigger performance improvement plans. The asymmetry is the bug. If the same team keeps producing burnout, the team is the incident. Treat it like one: timeline, contributing factors, action items that change intake, not slogans that change nobody.

Managers control more of this than they admit. Load needs honest accounting of sustainable hours, not heroics as the baseline. Low autonomy at high load is the fastest path to cynicism in Maslach’s model. Recognition is not stock; it is explicit acknowledgment that the work was hard. Week-to-week surprise is a tax on executive function. If “how are you” is performative, overload stays invisible until the exit interview.

AuDHD contributors often surface the failure earlier. Sensory load, ambiguous briefs, and meeting-heavy cultures cost more when nervous system and executive function are already under tax. Designing for that edge — optional meetings, written decisions, quieter async defaults — improves the median. The accommodation is an affordance. If only the people with the least spare capacity can survive your process, the process is the product defect.

Silence is expensive debt. Most cases I have seen had weeks of private awareness before public collapse. Silence persisted because honesty was career-coded as weakness. A 1:1 that always returns “fine” is not a feedback channel. It is theatre. Belief on the first overload signal, not the third, is the difference between a recoverable quarter and a resignation.

Managers owe written escalation when load exceeds capacity. Customer-visible date slips beat invisible overtime. If a project requires sustained extra hours to hit a committed date, that fact belongs in writing to leadership and, when appropriate, to the customer. It does not belong only in the bodies of the people doing the work.

Contributors owe early, written naming when gauges drift — sleep, irritability, loss of interest in problems they used to chase — to a manager who has earned trust. Individual boundaries help. They do not replace fixing intake. If no manager has earned that trust, the problem outgrows this essay. The environment is unsafe by design.³

Tradeoffs

Reduce load versus renegotiate roadmap. Cutting scope costs political capital with upstream stakeholders. It is still cheaper than rehiring and re-onboarding a senior engineer you trained for a year. Frozen headcount is not a reason to keep the same intake. It is a reason to cut the list.

Individual coping versus systemic fix. Sleep, boundaries, and early signals are real. Framing burnout as self-care alone lets the org off the hook for intake it controls. Do both. Do not let “have you tried yoga” close the ticket.

When “push through” is valid. Crunch with a defined end date, shared sacrifice, and post-ship recovery can work. Crunch as the permanent operating mode is extraction. If the recovery never arrives, it was not a sprint. It was the new baseline wearing a costume.

Close

The orgs that retained people through hard seasons treated capacity as a budget with a ledger, not a virtue test. They said no in writing. They celebrated shipping without automatically doubling the next assignment. Boring management. Also how you keep the engineers who remember where the bodies are buried.

Treat recurring burnout like recurring outages. Stop blaming the on-call human. Instrument the intake system. Fix the architecture. Write the runbook so the next rotation does not relearn it in tears.

— JV · Dark Heart Labs.

References

John Allspaw, The Art of Capacity Planning, and later writing on adaptive capacity in web operations. Utilization and headroom as measurable system properties — the same lens, applied to teams running hot for quarters. ↩
Christina Maslach and Michael P. Leiter, The Truth About Burnout (Jossey-Bass, 1997), and the Maslach Burnout Inventory research program. Six organizational risk factors as the standard diagnostic, environment over character. ↩
Christina Maslach, Wilmar B. Schaufeli, and Michael P. Leiter, “Job Burnout,” Annual Review of Psychology 52 (2001). The review article that put three decades of occupational evidence in one place, including knowledge work. ↩