[abewheel]

Public workflow assessment rebuilt retention review graphs and parsed the same governed Markdown for each schedule row. Share bounded intermediate calculations within one workflow or Program readiness invocation, then discard them so later calls observe edits. Nested calls retain verified snapshot behavior; validation, approval rules, model v13, and workflow contract 1 remain unchanged.

On the affected 462-record workspace, an unprofiled Node 24 assessment fell from 39.04 seconds to 1.42 seconds (27.6×). Including load and full record validation, elapsed time fell from 42.33 seconds to 4.47 seconds. The complete assessment JSON is byte-for-byte identical, including all 23 work items.

Validation: two independent combined code/security reviews clean; 43 focused tests; CI=true npm run validate on Node 24; both package dry runs; npm audit (zero vulnerabilities); fresh consumer validation (224 records, zero errors/warnings); real-browser Program Overview and Work Queue checks with no console errors or failed requests. Regression tests cover bounded review reads, next-call Markdown refresh, nested snapshot reuse, concurrent isolation, and failure cleanup.

Coordinates both packages for backward-compatible 0.18.7 publication. No customer record migration or lockfile upgrade is included.

Published FileGRC 0.18.7 from merged commit 849ad40. Trusted publication and installed README smoke passed; both npm packages expose 0.18.7 as latest with provenance.

Hosted runtime upgrade landed at dea3276cfac18c2bb9a7e140df20870f6e4558df. Hosted validation and deployment passed, including 124 browser tests. Production is running that commit and engine 0.18.7. Live verification returned all 23 work items without a 504, failed data request, or console error. Production server assessment took 6.32 seconds; complete fresh and subsequent reads took 21.3 and 13.5 seconds. Customer source records and lockfile remain unchanged.