rte-04 reproduction
Public source retained with the related article. Generated results, dependency directories, runtime storage, secrets, and oversized binary artifacts are intentionally excluded.
README
# RTE-04: persisted preparation and active identities
Run the public teaching model with Node.js:
```sh
node experiment.mjs
```
It uses only the standard library and in-memory maps. It writes no files,
contacts no services, and needs no credentials or application checkout.
The assertions distinguish publication prewarming, warm reads, a volatile-cache
flush, a new active dependency, corrupt cache/stored payloads, and a missing
required dependency. A deliberate same-ID content mutation then demonstrates
why identity-based reuse depends on a real versioning contract.
The expected final message is `All preparation-path assertions passed; no
latency was measured`. Operation counts describe this model, not the engine's
query counts or performance. JSON and maps simplify its representation; the
historical implementation had separate projection and context-bundle readers,
database persistence, codecs, and active-time queries. Concurrency, distributed
cache behavior, compression, pricing evaluation, and atomic publication are
outside this model. It contains no private article text or research history.
The model's JSON decoder checks its embedded identity and catches parse errors.
Those simplified checks are not an oracle for the application's actual decoder
or exception behavior. A model fallback must not be cited as proof that every
malformed runtime payload is automatically repaired.