CASCADS

Engineering evidence · published 9 August 2026

At-most-once social publishing under process contention.

We started 8 independent Node.js processes against the same publication key in each of 25 isolated rounds. The study observed exactly one provider callback per round.

Author: Victor LaybatsPublisher: IVRYNRuntime: v24.13.0 on linux x64Updated: 2026-08-14T00:48:44.265ZPermanent version: 2026-08-14T00-48-44-265Z
200independent process attempts
25isolated contention rounds
25provider callbacks observed
0duplicate provider callbacks

Result

Every round satisfied the study invariant. Across 200 simultaneous attempts, 25 workers reached the provider stub and 175 contenders were rejected before the side effect.

Measured invariant

For one content item and one network key, exactly one process may call the provider while the durable claim is active.

Observed outcome

25 total callbacks across 25 rounds, with 0 duplicate callbacks and 0 unexpected worker outcomes.

Method

  1. Create a new temporary ledger for each round.
  2. Start 8 independent Node.js processes and hold them at a common barrier.
  3. Release the barrier so every process contests the same item and platform key.
  4. Count callback marker files written only inside the provider stub.
  5. Require one delivered state, 7 blocked contenders and one callback in every round.

The tested source was lib/publication-claims.js at commit 582ac150398f27563f7eea30575fc5bb193a8364. Its SHA-256 was 26def4d026c9df1715e33ad9bc828261ec38daa60f44ef9d818253675ac043b5. Uncommitted changes to that source file: yes.

What this does and does not show

The result is narrow evidence about one local concurrency boundary. It is not evidence of customer outcomes, platform acceptance or distributed exactly-once delivery.

Limitations

  • This is a controlled single-host process-contention study, not a production traffic sample.
  • The provider callback is a local deterministic stub and no social-network API was called.
  • The result does not test multi-host or distributed filesystems.
  • The study covers the shared X and YouTube claim module through the X branch only.
  • A passing run is evidence for this implementation and environment, not a universal guarantee.

Data and reproducibility

The summary, all round-level results, schema and checksums are public. No customer content, account identifier or provider credential is present. The current page may be regenerated; the linked dated version is retained as the immutable citation target.

Benchmark JSONFull methods, environment and resultsRound-level CSVOne record per isolated roundJSON SchemaMachine-readable field contractManifestByte counts and per-file hashesSHA-256 checksumsIntegrity verificationReproduction noteRepository entrypoint and parametersCC BY 4.0 licenseScope and attribution termsGitHub mirrorPublic repository copy

The original compilation, authored method and generated benchmark results are licensed under CC BY 4.0. Third-party names, trademarks, APIs, software and facts are excluded.

Suggested citation

Laybats, Victor. “At-most-once social publishing: a process-contention study.” Cascads Research, IVRYN, 9 Aug. 2026. https://cascads.com/research/publication-proof/2026-08-14T00-48-44-265Z/