Skip to content

test(collector): bind saturation success to the seeded event - #6

Draft
dinakarjs wants to merge 1 commit into
test/collector-fault-harnessfrom
test/collector-saturation-seed-followup-20261108
Draft

test(collector): bind saturation success to the seeded event#6
dinakarjs wants to merge 1 commit into
test/collector-fault-harnessfrom
test/collector-saturation-seed-followup-20261108

Conversation

@dinakarjs

@dinakarjs dinakarjs commented Nov 8, 2026 •
edited
Loading

Copy link
Copy Markdown
Owner

Summary

The saturation observer accepted any nonempty proper subset of submitted events as committed, so a pressure event could replace the blocked seed and still pass. Require the first submitted event in both observed attempts and backend commits. Preserve allowed queue loss, extra pressure-event commits, duplicate attempts and observation reordering.

This is a separate three-file follow-up to agentrust-io#100, based on its approved head d70cdd9. It leaves that upstream PR and branch unchanged. Keep draft for later adoption after agentrust-io#100; do not merge this fork PR into the approved branch merely to transfer the patch.

Contract impact

  • No wire-contract change. Only experiment assertions, observer tests and documentation change.

Evidence

  • Eight observer tests pass normally and under python -O on Linux/Python 3.12.14, using opentelemetry-proto 1.45.1 and protobuf 7.36.2.
  • Substituting the original fidelity.py produces four assertion failures across the two new methods; original six methods remain passing. Restored implementation passes.
  • Positive controls allow extra pressure-event commits and duplicate/reordered attempts while retaining observed loss. Reversed input identifies its first event as the seed.
  • Fresh hosted Collector run 42506784329 passed on merge 07f41b6 containing head f80919f: 12 baseline cases, eight observer unit tests and eight stress cases. Both SDK saturation cases recorded 8 local evidence rows, 1 observed attempt and 1 commit, with the new seed identity assertion passing. Artifact 13628150246 was uploaded (68,537 bytes).
  • CI run 42506784373 passed all Python 3.10–3.13 and TypeScript jobs. Historical runs remain attributed to their original heads.

Privacy and security

  • Synthetic fixtures only; no credentials or production identifiers added.
  • Independent review remains pending.

Limits

Fresh execution covers the synthetic loopback pipeline only. This does not establish persistent queue recovery, exactly-once delivery or production reliability. No SDK runtime, dependency pins, workflow triggers or repository settings change.

Prepared with OpenAI Codex; DCO signed.

@dinakarjs
test(collector): require saturation seed observation and commit
f80919f 
Bind the saturation positive control to the first submitted event while preserving allowed loss, retries and observation reordering. Add substitution and missing-seed negative controls and retain historical pipeline evidence boundaries.

Prepared with OpenAI Codex.

Signed-off-by: Srinivasa Dinakar Jainti <dinakarjs@gmail.com>
@github-actions

github-actions Bot commented Nov 8, 2026 •
edited
Loading

Copy link
Copy Markdown

🟡 Contributor Check: MEDIUM

Check Result
Profile MEDIUM
Credential LOW
Overall MEDIUM

Automated check by AgenTrust Contributor Check.

@github-actions github-actions Bot added the needs-review:MEDIUM Contributor check flagged MEDIUM risk label Nov 8, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

needs-review:MEDIUM Contributor check flagged MEDIUM risk

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@dinakarjs