storage capacity: clarify pod scheduling

Michelle pointed out that late binding works a bit differently
(https://github.com/kubernetes/enhancements/pull/1353/files#r369707568).

The non-goal about prioritization is meant to explain that while this
would be possible, it's not planned to be implemented (yet).

keps/sig-storage/20191031-storage-capacity-constraints-for-pod-scheduling.md

@@ -97,11 +97,13 @@ provisioned and ask the CSI controller service to create the volume
 such that is usable by the node (via
 [`CreateVolumeRequest.accessibility_requirements`](https://kubernetes-csi.github.io/docs/topology.html)).
 
-If these volume operations fail, pod creation gets stuck. The
+If these volume operations fail, pod creation may get stuck. The
 operations will get retried and might eventually succeed, for example
-because storage capacity gets freed up or extended. What does not
-happen is that the pod is re-scheduled to some other node which has
-enough storage capacity.
+because storage capacity gets freed up or extended. A pod with an
+ephemeral volume will not get rescheduled to another node. A pod with
+a volume that uses delayed binding should get scheduled multiple times,
+but then might always land on the same node unless there are multiple
+nodes with equal priority.
 
 A new API for exposing storage capacity currently available via CSI
 drivers and a scheduler enhancement that uses this information will
@@ -135,6 +137,10 @@ reduce the risk of that happening.
   pending volume operations. This would depend on internal driver
   details that Kubernetes doesn’t have.
 
+* Nodes are not prioritized based on how much storage they have available.
+  This and a way to specify the policy for the prioritization might be
+  added later on in a separate KEP.
+
 * Because of that and also for other reasons (capacity changed via
   operations outside of Kubernetes, like creating or deleting volumes,
   or expanding the storage), it is expected that pod scheduling may