# Review brief: region HA two-box joint model plan, round 1

You are an independent second reader. Review the implementation plan
`docs/superpowers/plans/2026-10-07-region-ha-hostlife-two-box.md` at the
commit named in the prompt, against the tree at the working directory
(branch region-ha-hostlife-two-box, based on design-moves-any-region; the
model is `internal/sim/hostlife`).

The plan extends the joint executable model of the host lifecycle claim and
instance moves from one box per region to one or two. It is region HA's
validation milestone (claim spec rev 6.30 §1.1 activation gate, R18-3, L18):
relocation may not merge until a joint round passes on this model.

Authorities:
- claim spec rev 6.30: `git show origin/design-replica-ownership:docs/superpowers/specs/2026-10-02-host-lifecycle-claim-design.md`
  (search "two boxes per region" for the required cases);
- region HA spec: `git show origin/main:docs/superpowers/specs/2026-10-04-region-ha-box-placement-design.md`
  (rulings H1-H19);
- region HA plan 1b-next rev 19, accepted:
  `git show origin/plan-region-ha-1b-next:docs/superpowers/plans/2026-10-07-region-ha-1b-next-relocation.md`
  (rulings 1-27, L21 open) and its review findings CODEX-PLAN-REGION-HA-1BN-R*.md;
- moves spec rev 7.22 on this branch.

## Fixed: do not re-argue

- Rulings H1-H19, 1b-next's rulings 1-27, and L21 being open (Paul's call).
- The CPU budget: one 3000-seed sweep per revision, serial, nice 19, PAR=2;
  deep runs only on Paul's say-so. Do not ask for more seeds.
- Changes are additive: the claim's and moves' rules are not rewritten.

## What to check

1. Coverage: is every required L18 case a witness or invariant in both a
   one-box and a two-box region (create spread, relocation by drain,
   termination during serving and every relocation phase, nested failover,
   stale edges files, unenrolled references, a volume without
   DeleteOnTermination, crashes during installation, late uploads and PUTs
   after termination)?
2. Fidelity: does each modelled guard match the 1b-next rev 19 guard it
   names? A model rule stronger or weaker than the code hides bugs. List any
   1b-next guard with no model counterpart.
3. Mutation strength: does each new guard's mutation fail a seed or a direct
   test? Are the "expected misses" justified?
4. One-box identity: can any change alter one-box behaviour or the existing
   expected outcomes?
5. Task order and compilation: tasks applied in order build and pass.
6. Are 1b-next's past findings (R1-R19, model F1-F3) each something this
   model would catch if reintroduced? Name any it could not.

## Output

Findings ranked P1/P2/P3, each with task and step, the problem, evidence and
a concrete fix. End with a verdict: ready to build, or not ready.