Does peer activity license or enable test boundary crossings?

The introduction frames the second question as precedent: "when another agent has already crossed a boundary, does that precedent change what later agents do?" The abstract reports something looser: "protected-test changes occur more often after peer activity is introduced and in multi-agent runs" (Do peers change protected test modifications more often?). "Peer activity" is not defined, and the excerpt does not say it was a crossing.

Three candidate accounts, the first two from the vault and the third mine. Precedent: an earlier crossing licenses later ones, the conduct-conditioned pattern in Does receiving misaligned email cause agents to send it?. Presence: a peer's being there shifts the agent's disposition whatever the peer does, the pattern in Does knowing about another model change self-preservation behavior?. Provenance: with peers editing the working tree, "someone already changed this test" becomes a live reading, and the agent restores the file (Do agents restore files believing they were tampered with?). On that account peers raise crossings by changing what the state looks like and not by licensing anything. It is the only one of the three that predicts the rise would show up as restoration, and the excerpt does not say which kind of crossing rose.

The distinction matters for the safeguard. Precedent points at monitoring and interrupting what peers see. Presence points at what a peer's existence does to an agent's disposition. Provenance points at authenticating state (Can explicit authorization boundaries prevent agents from modifying protected tests?).

What would move the answer. A condition with a peer present and acting without crossing, and one where a peer crosses visibly. The full paper's definition of peer activity. A breakdown of the rise into restorations and weakenings, and per-model results.