Corollary: One selected design used on fresh copies

Fix $N$, payload dimension $d=2^{\ell }$, total full-diamond allowance $0<\epsilon <1$, and $0<p<1$. Use $N$ fresh copies of the same resource and port design. Select one device matrix $U$ certified to per-session full-diamond error $\epsilon /N$. Then the same decoder core can be used separately in all sessions, and the convoy's joint full-diamond error is at most $\epsilon$.

A sufficient per-session released dimension is

At first retained residual dimension $b_{\ast }$, $f=1$, it suffices to take

rounded upward to a power of two. The device-selection probability is at least $1-p$, not a fresh input-outcome probability for each ship.

Proof. Apply Theorem thm:threshold with per-session target $\epsilon /N$, and Theorem thm:convoy with $e_i=\epsilon /(2N)$. The same device property applies to every identically prepared copy; it need not be resampled.