The Loud Harbour Deterrence, Salience, and the Administrator who Finances its Own Backfire

A safe harbour must be administered, and administration is visible: publishing who is capped is publishing attribution. On the programme’s monitoring substrate, a companion paper shows that such publicity backfires beyond a critical loudness $\kappa_c$ — mild caps become worse than no harbour. This paper asks what loudness the administrator would choose. An exact identity makes the cost side degenerate: salience is a tax on the private production value of observation and on nothing else, so absent any reward the optimum is a corner — silence. The reward is built rather than assumed: a Becker-type awareness population in which exposure to enforcement deters, with the deterrence curve’s shape derived from the exposure process alone. Two closed-form thresholds organize the design space, each asking one number of the deterrence technology: caps tighter than $t_{\text{silent}}$ are administered in silence — loud harbours are for loose caps — and above a deterrence weight $\beta_c$ the administrator optimally selects a loudness inside the backfire band, financing its own backfire. Silence is priced at the origin; backfire is priced at the edge. Where marginal deterrence decays slowly, the silence boundary is only local, and silencing is discontinuous.