# Review brief: region HA plan 1a (foundations), round 1

You are an independent second reader. Review the implementation plan
`docs/superpowers/plans/2026-10-05-region-ha-1a-foundations.md` at the
commit named in the prompt, against the tree at the working directory
(Go, module `github.com/carlosframework/platform`) and the spec
`docs/superpowers/specs/2026-10-04-region-ha-box-placement-design.md`.

## Fixed: do not re-argue

- The spec's rulings H1 to H9.
- The split into 1a (no claim dependency, this plan) and 1b (after
  `internal/claim` lands). Do not ask 1a to wire anything into create,
  reconcile, the activator's restore path or relocation.
- The three departures the plan states at its top (content hash in place
  of S3 version ids; the base64 RSA identity signature; `snapshot` as a
  periodic full publish). Report only if one fails to keep the spec's
  property.

## What to check

1. **Does the code compile and do the tests pass as written?** Read each
   task's code against the real tree: every identifier, signature,
   import and file path it uses from existing code (`release.Store`,
   `registry.Timer`, `registry.Route`, `edgesFile`, `foldHomes`,
   `edgesFilePlacementView`, `views`, `markStale`, `testKeyFP`,
   `provision.InstanceID`, the ec2 SDK types, `unix.Inotify*`). Name the
   exact line that would not compile or the test that would fail, and why.
2. **Does each test actually pin the behaviour its name claims?** A test
   that passes for the wrong reason, or passes before the change.
3. **Correctness of the mechanisms**: the manifest's ordering under
   concurrent or late writes; `readStable` and symlink races; the inotify
   watcher (event parsing, lost events, new directories, close);
   `dialBeats` against every pre-existing home-map rule; `Verify` in
   `boxid` and `termfence`; the bucket policy statements (would any
   statement deny something that works today, especially lease releases
   or existing writers under `control/`?); the EventBridge and SQS tofu.
4. **Additivity**: anything that changes behaviour on a one-box region
   or breaks a reader of an existing wire shape.
5. **The Review Focus list**: is each of the five pinned by a test in
   the task that owns it?
6. **Task boundaries**: a task that cannot be built and tested alone, in
   order, with only the earlier tasks present.

## Output

Findings ranked P1 (would not compile, a test fails, or unsafe), P2
(wrong or missing), P3 (clarity). Each: task and step, the problem, the
exact evidence (file:line in the tree, or the plan's own line), and a
concrete fix. End with a verdict: ready to build, or not ready.