refactor: harden diagnostics and consolidate runtime code

audit/internal/platform/sat_test.go

@@ -273,6 +273,32 @@ func TestCheckNvidiaJobHealthReturnsErrorForSelectedResetRequiredGPU(t *testing.
 	}
 }
 
+func TestCheckNvidiaJobHealthReturnsErrorForSelectedNVLinkDegradedGPU(t *testing.T) {
+	oldExecCommand := satExecCommand
+	satExecCommand = func(name string, args ...string) *exec.Cmd {
+		switch name {
+		case "nvidia-smi":
+			return exec.Command("sh", "-c", "printf '0, NVIDIA H100 PCIe, 38, 46.89, 0, 0, 81559\n1, NVIDIA H100 PCIe, 39, 47.00, 0, 0, 81559\n'")
+		case "dmesg":
+			return exec.Command("sh", "-c", "printf 'NVRM: knvlinkSetDegradedMode_IMPL: GPU1 marked Degraded. Error originated on linkId 1!\n'")
+		default:
+			return exec.Command(name, args...)
+		}
+	}
+	t.Cleanup(func() { satExecCommand = oldExecCommand })
+
+	msg, err := checkNvidiaJobHealth([]int{1})
+	if err == nil {
+		t.Fatal("expected NVLink degraded health check error")
+	}
+	if !strings.Contains(msg, "GPU1 NVLink degraded mode (linkId 1)") {
+		t.Fatalf("unexpected message: %q", msg)
+	}
+	if msg, err := checkNvidiaJobHealth([]int{0}); err != nil || msg != "" {
+		t.Fatalf("unaffected selected GPU0 got msg=%q err=%v", msg, err)
+	}
+}
+
 func TestWriteNvidiaGPUStatusFilesCreatesPerGPUFiles(t *testing.T) {
 	dir := t.TempDir()
 	oldExecCommand := satExecCommand