ELX Scoring Rev 2.2 + Session Feedback rebuild (parts 1–5) — qa twin
- fix(scoring): give unscoreable pending rows a terminal state so the rescore sweep converges
- feat(scoring): ELX Scoring Rev 2.2 part 1 — move all arithmetic out of the model
- feat(scoring): ELX Scoring Rev 2.2 part 2 — lexicon lock + plain-English feedback writer
- feat(practice): ELX Scoring Rev 2.2 part 3 — Session Complete screen rebuilt
- feat(practice): ELX Scoring Rev 2.2 part 4 — persona agent time-aware exit + anti-capitulation
- feat(scoring): ELX Scoring Rev 2.2 part 5 — calibration runner + compact evaluation payload
- docs(practice): ELX Scoring Rev 2.2 engineering notes
- fix(scoring): ELX Rev 2.2 calibration tuning — lenient writer JSON, tighter override tests
- Merge pull request #511 from tpromni/feat/elx-scoring-rev-2-2-qa
Source: tpromni/coach-pulse-consumer-ui → qa