In Praise of Maintenance

In April 1979, a pump failed three floors beneath the county records office, and for eleven days nobody noticed. The water rose in silence. By the time a clerk smelled the damp, four decades of council minutes had become pulp. What followed was not a dramatic collapse but a long, modest accounting: inventories, apologies, dehumidifiers, a committee. Most of the people who saved what remained never appear in any history of the building. This essay is about them — and about the kind of work they do, the kind that succeeds by nothing happening.

01The work beneath the work

Maintenance is easy to underestimate because it is designed to be invisible. When it succeeds, there is no event: no headline, no ribbon, no screenshot. The pipe holds, the index rebuilds, the night shift hands over a clean log. Institutions run on a steady diet of this unremarkable competence, yet budget meetings treat it as residue — whatever is left once the interesting spending is done. The asymmetry is strange. We will happily fund the thing that breaks spectacularly, because failure at least produces a story, but the quiet work of prevention has no story, and so it gets deferred. Deferred, note, not avoided. Every maintenance hour skipped is a loan taken out against a future shift, at interest.

02Failure is a process

Serious accidents are almost never a single moment. They are a slow negotiation between small compromises, each one reasonable on the day. A sensor runs flaky but stays “within tolerance.” A valve is serviced in spring instead of winter because spring is only months away. Look closely at any major failure and you find not a villain but a schedule — a long list of minor decisions that accumulated into a catastrophe no single decision ever intended. That is why post-mortems that end with a person’s name have failed before they begin. The useful question is never who missed the sign; it is why the sign was so easy to miss, and who, if anyone, was resourced to look for it.

03A charter for maintainers

None of this means we stop building new things. It means we budget for the decades after the launch party. Three commitments would change the shape of the work. First, name the maintainers: the person who answers at 2 a.m. belongs in the story of the system, not only in its incident log. Second, measure drift, not just uptime — a healthy system is one whose small problems are found early, spoken about plainly, and fixed in daylight. Third, treat deferred maintenance as debt on the books, because that is precisely what it is. The pump beneath the archives failed quietly. The question for any institution is whether anyone is listening for that particular silence.