TML-2984: slice 06 close-out — SDoD walked

Signed-off-by: Serhii Tatarintsev <tatarintsev@prisma.io>

Author: SevInf

projects/lsp-interpreter-diagnostics/plans/plan.md

@@ -233,7 +233,7 @@ interpret stage.
 
 ### Implement M5: Config-failure surfacing + last-good retention
 
-**Status:** ► In progress — slice 06, branch `tml-2984-slice-06-config-failure`
+**Status:** ► In progress — slice 06 delivered (`353e5d4e7`, 4/4 SDoD), PR open; complete on merge
 
 _Outcomes_
 A throwing `loadConfig`/`createControlStack` surfaces as a push diagnostic on the

projects/lsp-interpreter-diagnostics/slices/06-config-failure/spec.md

@@ -43,18 +43,29 @@ either a diagnostic for a state we still destroy, or retention nobody can observ
 
 ## Slice Definition of Done (beyond CI / reviewer / project-DoD)
 
-- [ ] SDoD1 (TC-13) — first-load failure: push diagnostic on the config-file URI at
-      (0,0)–(0,1) with the error message; published even when the client supports
-      pull diagnostics; exactly once per failed load with multiple awaiting
-      documents; next successful load clears it (empty publish).
-- [ ] SDoD2 (TC-14) — reload failure with a previously loaded project: last-good
-      project keeps serving (schema diagnostics — including interpreter findings —
-      still returned on pull/push), config diagnostic alongside; a subsequent
-      successful reload swaps in the new project and clears the config diagnostic.
-- [ ] SDoD3 — first-load failure keeps today's no-project behavior (documents
-      unmanaged; existing regression pins hold); success path behavior-identical
-      (existing server suite untouched and green).
-- [ ] SDoD4 — zero new casts; no new dependency edges.
+- [x] SDoD1 (TC-13) — push diagnostic on the config URI at (0,0)–(0,1)
+      (`PRISMA_NEXT_CONFIG_LOAD_FAILED`, error message); exactly-once is structural
+      (single shared load promise); published for pull-capable clients; cleared on
+      next successful load. ✓ `353e5d4e7`
+- [x] SDoD2 (TC-14) — break→retain→fix cycle pinned: last-good serves full schema
+      diagnostics incl. interpreter findings, config diagnostic alongside;
+      successful reload swaps + clears. Reload-failure resolves with last-good
+      (funnels untouched). ✓
+- [x] SDoD3 — first-load failure = today's behavior (217 existing tests untouched;
+      1 test re-pinned by charter — it encoded the destroy-on-reload behavior this
+      slice replaces; reviewer confirmed everything it legitimately protected
+      survives in other pins). ✓
+- [x] SDoD4 — zero casts; no new edges; `lint:framework-vocabulary` at threshold. ✓
+
+**Slice-close ritual (2026-07-14):** single dispatch SATISFIED R1, zero findings;
+4/4 SDoD PASS; clear-site relocated from the brief's letter (`stopManagingProject` —
+whose only callers are the failure funnels) to the genuine stops-caring seam
+(`dropProjectWithoutManagedDocuments`), reviewer-blessed. Accepted residual for M6
+QA: after a failed *first* load with all schema docs closed, the marker persists
+until a doc reopens (lazy lifecycle by design) — QA script must exercise break→fix
+with and without open documents. Pre-existing superseded-failed-awaiter hazard
+recorded for backlog (blast radius narrowed by this slice). Manual QA: config
+break/fix cycle is in the M6 playground script by project DoD.
 
 ## Edge cases (pre-investigated)
 

projects/lsp-interpreter-diagnostics/trace.jsonl

@@ -111,3 +111,5 @@
 {"event_id":"102ba70b-2815-45a9-a9ff-f608e63d6248","schema_version":"1","ts":"2026-07-14T16:14:10.156Z","project_run_id":"lsp-interpreter-diagnostics","orchestrator_agent_id":null,"event_type":"dispatch-start","dispatch_id":"5969f7c0-14c6-4f68-8df6-2ed8453710a8","dispatch_name":"S6-D1 config-failure surfacing + last-good retention","subagent_type":"spawn_agent","model":null,"parent_dispatch_id":"af03de09-e8b6-4379-918c-581e95e0e241"}
 {"event_id":"c0c6ea09-f9f2-40f5-be24-918140380ad7","schema_version":"1","ts":"2026-07-14T16:14:11.263Z","project_run_id":"lsp-interpreter-diagnostics","orchestrator_agent_id":null,"event_type":"round-start","dispatch_id":"5969f7c0-14c6-4f68-8df6-2ed8453710a8","round_id":"6057fa81-b057-4313-b10b-b4f5790392b7","round_number":1}
 {"event_id":"70a7fde7-8b4e-468b-aaa3-f22b53558065","schema_version":"1","ts":"2026-07-14T16:14:12.425Z","project_run_id":"lsp-interpreter-diagnostics","orchestrator_agent_id":null,"event_type":"brief-issued","dispatch_id":"5969f7c0-14c6-4f68-8df6-2ed8453710a8","round_id":"6057fa81-b057-4313-b10b-b4f5790392b7","brief_byte_length":3910,"brief_content_hash":"6327b1c6383dfe8e11d0cd2d8f182c4c00a675eed8b0155e89e113a8f5eff765","brief_disposition":"initial"}
+{"event_id":"204d8a41-f963-44ff-a73a-177ca1a43f38","schema_version":"1","ts":"2026-07-14T16:44:14.378Z","project_run_id":"lsp-interpreter-diagnostics","orchestrator_agent_id":null,"event_type":"round-end","dispatch_id":"5969f7c0-14c6-4f68-8df6-2ed8453710a8","round_id":"6057fa81-b057-4313-b10b-b4f5790392b7","verdict":"satisfied","findings_filed":0,"wall_clock_ms":4500000}
+{"event_id":"22388532-a5fe-4089-bb81-aeb0cf6f1e38","schema_version":"1","ts":"2026-07-14T16:44:14.851Z","project_run_id":"lsp-interpreter-diagnostics","orchestrator_agent_id":null,"event_type":"dispatch-end","dispatch_id":"5969f7c0-14c6-4f68-8df6-2ed8453710a8","result":"completed","wall_clock_ms":4500000}