lar-03 reproduction

Public source retained with the related article. Generated results, dependency directories, runtime storage, secrets, and oversized binary artifacts are intentionally excluded.

README

# LAR-03 worker-throughput experiment

This synthetic Laravel 13.20 application separates logical operations,
queue attempts, and completed projection effects before comparing worker
counts. A pinned Redis container owns queue coordination while an isolated
SQLite WAL database owns application effects and the deliberately serialized
hot row. This keeps broker coordination separate from application database
pressure.

The full runner recreates all retained evidence, checks the report, rejects 20
semantic mutations, audits locked dependencies, retrieves primary sources,
and verifies raw-log and source hashes:

```text
./run.sh
```

The drafting gate remains closed until this runner and a clean committed export
both pass. `smoke.sh` remains the narrow one-job diagnostic when the full
matrix is unnecessary.

Retained files