CrossDecoy

The finite-decoy estimator reduced to the infinite-decoy ideal

3/3 passed in 49ms

Test	What it does	Result
test_decoy_limit	Halving nu1 and nu2 halves both deviations from the ideal to atol 0.05, second order leaving 2.023 at nu1 = 0.024, nu2 a fixed fraction of nu1 so the full two-decoy expression runs.	✅ pass
test_decoy_sandwich	Over 4000 random self-consistent channels -- four decades of transmittance and background, misalignment to 5%, decoys to nu1 + nu2 = 1.71 mu -- the Y1 lower bound never exceeds the infinite-decoy Y1 and the e1 upper bound never falls below it, at zero slack.	✅ pass
test_decoy_vacuum	At a fixed signal and first decoy nu2 = 0 dominates every positive nu2 on both sides over 2000 random channels at zero slack: weaker decoys, not more intensities, converge onto the ideal.	✅ pass