Task · SKT-0172

Preserve fractional Solace publish and drain capacity across observation windows

Description

The Solace queue model rounds publishRateloadwindowSeconds and consumeRate*windowSeconds independently each tick. At normal 60-second cadence a plausible sparse flow below half a message per window disappears forever, and fractional drain capacity also vanishes or is inflated. Preserve realistic configured rates; no metric names, labels or types change.

Acceptance Criteria

Definition of Done

Implementation Plan

Read existing queue step and tests; prove sparse publish/drain failures via Tick; retain bounded fractional work across windows without idle delivery banking; scoped remote gate, inventory comparison and dsreview, then composed root landing.

Implementation Notes

Landed task 3d1e3d6 in exact green merged30e154c/tree1290aa3 preserving foreign main. Root recomputed accepted patch hash98a8af49. Actual Tick red/green proves fractional publish/drain and conservation/rollups/no idle bank/pause; selected CLI80/80 names and labels retained. CodeRabbit7/7 composed files zero findings, full remote gate20261023T095319Z-synthkit-94a6d21e exit0 including race, local gzip-aware2003 scan clean. No public schema inputs, generator not applicable. Hosted CI watched separately; existing custody aggregate timeout remains separate repair.

Final Summary

Integral queue traffic now carries bounded sub-message publish and active-backlog service work across windows instead of rounding sparse work away; empty queues expire spare service. Rates and public telemetry surface unchanged.

View the source file on GitHub