Operational Sustainability: Running Well for Years

Why this matters

A platform runs brilliantly for eighteen months on its builders' tribal knowledge, then the builders move on and every incident becomes archaeology: undocumented flags, unowned jobs, runbooks describing a system two redesigns old. The new team inherits toil, not a platform. Sustainability is the difference between a system that runs and a system that can be run by others; the second needs deliberate design from the first year, not a documentation sprint in the last month.

Concepts

Toil is operational work that is manual, repetitive and automatable; the SRE discipline caps it (a fraction of time, tracked, with excess converted to engineering). Track toil per team per quarter: pages, manual deploys, ticket-driven changes. Anything recurring gets automated, eliminated or accepted explicitly with its cost named. Toil accepted silently grows until it is the job; toil budgeted visibly shrinks because it competes with features honestly.

Knowledge stays alive through ownership and rehearsal. Every runbook has an owner and a rehearsal date; stale runbooks fail loudly in drills (M23's handover audit checks this). Documentation carries expiry: architecture notes, contact lists, dependency maps reviewed on schedule, with the review itself logged. Docs without owners rot at the speed of organizational change; expiry makes the rot visible before it matters.

Handover is a practice, not an event. The L69-adjacent standard from M23 applies inside this lesson: the package (runbooks, ownership map, known issues, pending work, contact paths) lets a stranger operate the system next week. Test handovers with rotations before departures force them: the successor operates solo while the predecessor watches, silently, for a week. Gaps found this way are gifts, found early.

Worked example

A fictional service prepares its sustainability review: toil measured at a tracked fraction with the top item automated this quarter, three runbooks rehearsed with dates, docs within expiry, and a handover package that a peer uses to run a drill solo. The review quotes each evidence; the gaps become next quarter's owned work, not shame.

Common wrong move

The documentation sprint before a departure. Three weeks of frantic writing produces volume without rehearsal, and the successor inherits prose instead of practice. Little, owned, rehearsed and current beats big, orphaned and stale every time.