What caused the decline in DSE wiki retained writes?

The conclusion says: "Our wiki reconstruction separates the decline in retained writes from later cleanup without establishing its cause." The reconstruction is the paper's own work on the public-wiki case, in which most activity appeared on DSE wiki, a public wiki within the ProWiki service. It found two things: retained writes declined, and cleanup came later. The paper reports that the decline is not the cleanup. It does not say what the decline was.

Two candidate causes, both mine and neither in the excerpt. Agents may have written less, having finished, stopped or changed what they did. Or the writes may have been removed by some means other than the later cleanup, so that what was retained fell while what was written did not. "Retained" implies that some writes were not kept, which makes the second reading possible, but the excerpt does not say which happened.

It matters for the paper's own design. The coordination episode links "response history" to transfers and task authority (Should defence units span multiple executions and agents?). If a decline in activity is read as a sign that a response worked, the reading needs the cause: an agent-side change, an operator's action, or an outside editor's. The proposed test of "recurrence after channel closure and state quarantine" also presumes the effect of a response can be told from a decline that would have happened anyway (How do we tell coordination apart from shared causes?).

The paper is candid that the cause is not established, and the note keeps that limit: the decline is a reconstruction result with an open cause, not a finding about agents.

What the excerpt does not give. The size and timing of the decline, when the cleanup happened or who did it, what "retained" means in the wiki's terms, and any candidate cause.