BREAK A — withdrawal asymmetry:
  w_max=3%: dividend stays full until inflationary drift = 3%, then it pays for the anchor.
  w_max=6%: dividend stays full until inflationary drift = 6%, then it pays for the anchor.
  -> Independence is ASYMMETRIC: free when KI ADDS (deflationary drift, the CS design case),
     broken when KI must SUBTRACT beyond its withdrawal capacity (inflationary drift/shock).

BREAK B — shared issuance ceiling (C=4%): dividend full until pi*=1%, then the corridor crowds it.
  -> This bites the CORRIDOR, not zero: a positive anchor competes with the dividend under a cap; zero leaves max room.

BREAK C — real cost of inflation (a=20): real dividend at pi*=0/+3% = 1.000 / 0.982 of base.
  -> 'Dividend independent of pi*' is only approximate; the dividend is MAXIMIZED at zero. Also reinforces zero.

VERDICT: one real break (A), two that reinforce zero (B,C).
  The decoupling holds in CS's design regime (deflationary drift, anchor held by ADDING money).
  It fails only when holding the anchor requires net CONTRACTION beyond KI's independent withdrawal
  instrument. Fix: the spec must NAME that instrument (bond/tax sterilization) so KI never has to
  reach into the dividend. With it, independence holds both directions; without it, only upward.

saved cs_independence_redteam.png
