# After evidence disposition

| Proof | Observed | Limits |
|---|---|---|
| R1 | Supplied primary fresh UI registration, two changed-port reconnects, stable saved tuple/ID/environment and1 route; video inspected and uploaded | Primary video backend15477; baseline-matched visual phase backend15474. No raw argv or credential dump |
| R2 | Actual blank discovered selection; config/listener24570→24571; savedalias/alias/null/null and ID stable; real descriptor success at both ports, routecount2 stable | Primary remains preferred active route; discovered target is independently exercised through existing desktop IPC |
| R3 username-only | Actual keyboard suggestion selection with typedubuntu, emptyport; configured nonexistentusername. Savedubuntu/null; effectiveubuntu/24571 before/after changing wrongusername again | No second valid account campaign |
| R3 port-only | Actual keyboard selection with emptyusername, typed24571; config24579 then24578. Savednull/24571; effectiveubuntu/24571 after explicit disconnect/reconnect | Configuredusername remainsubuntu; R2 covers blank username persistence |
| R3 both | Actual keyboard selection typedubuntu/24571; both configdefaults deliberately wrong, then changed wrong again. Saved/effective typedvalues and same IDs after reconnect | Desktop IPC runtime observation is not claimed as UI selection |
| V2 After | Four Electron1440/390×900 light/dark recordings, fields before selection and resulting labels, keyboard plus pointer | Representative case per cell; not both blank and explicit cases in everycell |
| V1 After | Fresh1440light supplied primary; supplemental existing-alias Add environment/dedup recordings per captured cell | Supplemental cell status and duration in inventory; no independent fresh registration per cell |
| Before V1 |1440light only | Three missing Before cells remain UNVERIFIED |
| Before V2 | Four supplied recordings | Baseline pair/settlement gaps and reduced-case coverage remain |
| Pixel comparisons | Only samequery/sole-suggestion control crops, four matching size/theme cells. Zero-delta8× and >16/255 masks, explicit counts | No full-frame/dialog-wide comparison because background/translucency context differs. No visual-regression pass inferred from thresholds |

Exact provenance/SHA columns, narrated UTC actions, artifact links, inspection/settlement data and final fixture state are in INDEX.md, PROVENANCE.md, pixel-analysis.json, inventory.json and FINAL-STATE.md. No source/forge/lifecycle changes or tests run in this phase.

Final: V1 supplemental4/4 and V2 representative4/4 recordings completed and inspected. Strict settlement15/24 pairs;9 failures (caret/toast), including all4 V1 result pairs. Before V1 remains1/4. No full-matrix acceptance claim. R1/R2/R3 runtime evidence is affirmative within the stated limits.