# #292 residual E2E fixture report

Date: 2026-08-10
Base: `7dc9f8ae` (#668 merged main)
Source commit: `7adaa1d7d3c05e88be392ca8b9e07dee85a67e26`
Scope: test harness only; no production/server/runtime behavior changed.

## Witnessed red before this source

Raw full-regression output from GitHub run `31338780878` after #668:

- Base E2E: 136 pass / 0 fail
- V3 Networks: 25 pass / 1 fail (`A sees B's network!`)
- Loop runtime: crashed before its Results trailer (`alias_not_found`)
- Loop npm-pack: 10 pass / 2 fail (`alias_not_found`)
- Loop self-mgmt: 2 pass / 3 fail (`alias_identity_mismatch`)
- Reported total: 214 pass / 7 fail

The network test had used the first registered global admin as an ordinary
tenant. The loop tests had reused the registration node token for different
node aliases. Current Hub identity checks correctly reject those fixtures.
No production authorization or identity gate was loosened.

## Exact-source Docker evidence

Low-disk derivative build (production artifacts inherited from the already
reviewed #668 exact image; this source changes only four copied test scripts
plus its test runner):

```text
tag=anet-test292-residual:dev
image=sha256:4c73933d56451f7db7ee3c3abcdfdaed8375eb27a38335347e0b36fee69cfdc5
TEST292_RESIDUAL_SOURCE_COMMIT=7adaa1d7d3c05e88be392ca8b9e07dee85a67e26
base-image=anet-test292-identity:dev
base-image-id=sha256:2e8938630aef1b4cf31911d10d324fe82d20c2f9d350c08b29ee8c16ebafe8c8
```

Command:

```sh
sg docker -c 'docker run --rm anet-test292-residual:dev'
```

Result:

```text
PASS: networks: Results: 26 passed, 0 failed
PASS: npm-pack: Results: 12 passed, 0 failed
PASS: self-mgmt: Results: 47 passed, 0 failed
PASS: runtime: Results: 21 passed, 0 failed
PASS: mutation: restoring admin A turns tenant isolation red
PASS: mutation: runtime exact-alias identity is load-bearing
PASS: mutation: npm-pack exact-alias identity is load-bearing
PASS: mutation: self-mgmt exact-alias identity is load-bearing
RESULT: PASS=8 FAIL=0
```

The real 60-second multi-cycle scheduler windows were retained; they were not
shortened for the focused run.

## Remaining #292 gate

The GitHub workflow still captures `tee`'s exit code for
`docker run ... | tee` and remains report-only. After this fixture candidate
merges, a separate narrow follow-up must capture `${PIPESTATUS[0]}` and prove
the exact full-regression image exits zero before #292 can close.
