diff --git a/docs/evidence/gen2-core-m5-local-medium-remediation-2026-08-30.json b/docs/evidence/gen2-core-m5-local-medium-remediation-2026-08-30.json
index 9c4b319..49e2708 100644
--- a/docs/evidence/gen2-core-m5-local-medium-remediation-2026-08-30.json
+++ b/docs/evidence/gen2-core-m5-local-medium-remediation-2026-08-30.json
@@ -2,7 +2,7 @@
   "evidence_id": "gen2-core-m5-local-medium-remediation-2026-08-30",
   "task_id": "GEN2-REM-CORE-M5-LOCAL-MEDIUM-001",
   "status": "CORE_ACCEPTED_UI_DEFERRED",
-  "source_revision": "5c95916648a78d25752c70881ae5c16f93dcdc27",
+  "source_revision": "8aa5179d47aef367ff7ef3732ed0ac603c5ffc89",
   "root_cause": {
     "discovery": "The approved llm-local gateway /v1/models endpoint performs SQLite-backed route updates while serving discovery. Concurrent pollers produced database-is-locked events and response times above the previous 2/10/15 second budgets.",
     "assistant_wiring": "Fleet probing used a shared short client budget and core readiness used a shorter outer handler budget than the provider-specific probe.",
@@ -11,7 +11,7 @@
   "changes": [
     "Provider-specific MEDIUM discovery and invocation budgets are bounded at 120 seconds.",
     "Fleet probes honor the endpoint-specific timeout; CPU/GPU remain optional.",
-    "Core readiness has an outer budget longer than the cognitive worker readiness probe.",
+    "Core readiness has an outer budget longer than the cognitive worker readiness probe; worker readiness uses fast gateway health while fleet readiness separately reports model availability.",
     "No schema, migration, privilege, owner, adapter, or architecture change."
   ],
   "production": {