diff --git a/csi.proto b/csi.proto
index 81627dd6..479c9f2a 100644
--- a/csi.proto
+++ b/csi.proto
@@ -1328,7 +1328,10 @@ message VolumeHealth {
   option (alpha_message) = true;
 
   enum Condition {
-    // Volume health condition is unknown; treat it as healthy.
+    // Volume health condition is unknown; If this is the initial
+    // report of volume health, treat it as healthy. If there was
+    // already a condition reported previously, treat unknown as
+    // unchanged from the previous condition.
     UNKNOWN = 0;
     // Volume is accessible and no problems are detected. 
     HEALTHY = 1;
@@ -1337,13 +1340,15 @@ message VolumeHealth {
     // maintenance.
     TEMPORARY_INACCESSIBLE = 2;
     // Volume is accessible currently, but is degraded. Data loss
-    // is unlikely as the storage system is fixing the problem
-    // automatically.
+    // is unlikely as the storage system is rebuilding to fix the
+    // problem automatically. For example, a replica is lost on
+    // the storage system but a spare is configured, and the storage
+    // system is resyncing/reconstructing.
     TEMPORARY_DEGRADED = 3;
     // Volume is accessible currently, but a problem is detected that
     // could lead to eventual data loss. Examples: recoverable hardware
-    // failure, cooling failure, equipment significantly past end of
-    // life.
+    // failure (i.e. the disk needs to be manually replaced by a human),
+    // cooling failure, equipment significantly past end of life.
     FAILURE_LIKELY = 4;
     // Volume is not accessible currently, and there is reason to
     // believe the outage may be permanent. Examples: unrecoverable
diff --git a/lib/go/csi/csi.pb.go b/lib/go/csi/csi.pb.go
index 3e34f4cb..1a8f0831 100644
--- a/lib/go/csi/csi.pb.go
+++ b/lib/go/csi/csi.pb.go
@@ -288,7 +288,10 @@ func (VolumeUsage_Unit) EnumDescriptor() ([]byte, []int) {
 type VolumeHealth_Condition int32
 
 const (
-	// Volume health condition is unknown; treat it as healthy.
+	// Volume health condition is unknown; If this is the initial
+	// report of volume health, treat it as healthy. If there was
+	// already a condition reported previously, treat unknown as
+	// unchanged from the previous condition.
 	VolumeHealth_UNKNOWN VolumeHealth_Condition = 0
 	// Volume is accessible and no problems are detected.
 	VolumeHealth_HEALTHY VolumeHealth_Condition = 1
@@ -297,13 +300,15 @@ const (
 	// maintenance.
 	VolumeHealth_TEMPORARY_INACCESSIBLE VolumeHealth_Condition = 2
 	// Volume is accessible currently, but is degraded. Data loss
-	// is unlikely as the storage system is fixing the problem
-	// automatically.
+	// is unlikely as the storage system is rebuilding to fix the
+	// problem automatically. For example, a replica is lost on
+	// the storage system but a spare is configured, and the storage
+	// system is resyncing/reconstructing.
 	VolumeHealth_TEMPORARY_DEGRADED VolumeHealth_Condition = 3
 	// Volume is accessible currently, but a problem is detected that
 	// could lead to eventual data loss. Examples: recoverable hardware
-	// failure, cooling failure, equipment significantly past end of
-	// life.
+	// failure (i.e. the disk needs to be manually replaced by a human),
+	// cooling failure, equipment significantly past end of life.
 	VolumeHealth_FAILURE_LIKELY VolumeHealth_Condition = 4
 	// Volume is not accessible currently, and there is reason to
 	// believe the outage may be permanent. Examples: unrecoverable
diff --git a/spec.md b/spec.md
index 940458b7..4af8c032 100644
--- a/spec.md
+++ b/spec.md
@@ -2336,7 +2336,10 @@ message VolumeHealth {
   option (alpha_message) = true;
 
   enum Condition {
-    // Volume health condition is unknown; treat it as healthy.
+    // Volume health condition is unknown; If this is the initial
+    // report of volume health, treat it as healthy. If there was
+    // already a condition reported previously, treat unknown as
+    // unchanged from the previous condition.
     UNKNOWN = 0;
     // Volume is accessible and no problems are detected. 
     HEALTHY = 1;
@@ -2345,13 +2348,15 @@ message VolumeHealth {
     // maintenance.
     TEMPORARY_INACCESSIBLE = 2;
     // Volume is accessible currently, but is degraded. Data loss
-    // is unlikely as the storage system is fixing the problem
-    // automatically.
+    // is unlikely as the storage system is rebuilding to fix the
+    // problem automatically. For example, a replica is lost on
+    // the storage system but a spare is configured, and the storage
+    // system is resyncing/reconstructing.
     TEMPORARY_DEGRADED = 3;
     // Volume is accessible currently, but a problem is detected that
     // could lead to eventual data loss. Examples: recoverable hardware
-    // failure, cooling failure, equipment significantly past end of
-    // life.
+    // failure (i.e. the disk needs to be manually replaced by a human),
+    // cooling failure, equipment significantly past end of life.
     FAILURE_LIKELY = 4;
     // Volume is not accessible currently, and there is reason to
     // believe the outage may be permanent. Examples: unrecoverable