TEST 1 — same engine, only KI's setpoint moves
  realized inflation (mean):  isolated -0.010   interop +0.000   corridor +0.029
  -> KI setpoint sets the level exactly.
  real K3 (cash dividend) identical across all three setpoints? True
  real K2 (stake) identical across all three setpoints?        True
  -> distribution is untouched by the inflation knob. Holding zero costs nothing on the dividend.

TEST 2 — CS (two instruments) vs bundled (one instrument): can you hold distribution while cutting inflation?
  CS real K3 at pi*=-1%, 0%, +3%:      0.049, 0.049, 0.049  (flat)
  bundled injection at pi*=-1%,0%,+3%:  0.049, 0.049, 0.098  (falls as you cut inflation)
  -> CS decouples the two; a bundled system must trade distribution against inflation.

TEST 3 — two members interoperate via EQUA at the common-zero anchor
  common-zero: bilateral rate end = 0.374  (reflects productivity gap 0.025; clean)
  B runs +3% corridor (sticky wages): rate end = 0.353  -> distortion -5.7%
  -> the external-layer result reappears INSIDE the integrated model: corridor distorts, zero doesn't.

TEST 4 — onboarding a high-inflation legacy member via KI_T, without starving distribution
  inflation: yr0 19%  yr4 11.6%  yr8 1.7%  yr12 0.04%  -> converges to anchor
  real K3 dividend continuous through disinflation? True  (grows with real output throughout)

saved cs_channel_test.png
