Ticket graphs

STOP brake for #449 (forge evolution Part 2, v1). DEFECT (mine, in the spec): #449 section 2 instructed the worker to read omni.git's refs/archives/<TS>/<repo>/heads/<b> to report 'ingest stamps' as a second clock. That is the exact confusion the operator called out: treating omni-git as A GIT rather than as A TOOL MADE OF GITS, and then conflating it with forge. MEASURED — omni is not a git: git -C omni.git rev-parse --abbrev-ref HEAD -> "HEAD" (the literal string) git -C omni.git ls-tree HEAD -> fatal: Not a valid object name HEAD git -C omni.git log -1 -> fatal: your current branch 'master' has no commit MEASURED — omni does not cover forge: 264 forge bare repos vs 145 repos in omni, only 29 overlap. omni covers 11% and MISSES 235 forge repos. It has ZERO refs for the jail lineage the ticket's own worked example depends on. (The '~86%' in the ticketer is BLOB coverage in `occ`, not repo coverage.) So the requirement was wrong three ways: architecturally (archive treated as a repo), empirically (89% miss), and pedagogically — it would have taught the next agent the same confusion. SUPERSEDED BY #452, which BLACK-BOXES omni: do not open it, do not walk refs/archives/*, do not use it for dates or provenance. config/forge/remotes/*.git is the only authority. The temporal dimension survives intact — git already carries every date needed (root/fork-point/tip + fork width), proven on the jail lineage: root 2026-07-08 20:38 -> fork 2026-07-11 02:23 -> jail tip 2026-07-12 14:07. DO NOT re-dispatch #449.

— browse congruency source (forge-parked tree)

— #449 Forge evolution PART 2 of 2: EVOLUTION TIMELINE (root -> fork-point -> tip dates + fork width, plus omni ingest stamps as a second clock) + mechanical CONTAINMENT retirement gate + the 118-blob orphan/merge-home worklist. Read-only; consumes #445's JSON. (stops · ACTIVE)