diff --git a/doc/src/sgml/ref/insert.sgml b/doc/src/sgml/ref/insert.sgml
index 1aa7a7ecbac..1c358b984ae 100644
--- a/doc/src/sgml/ref/insert.sgml
+++ b/doc/src/sgml/ref/insert.sgml
@@ -556,6 +556,15 @@ INSERT INTO <replaceable class="parameter">table_name</replaceable> [ AS <replac
     </para>
    </tip>
 
+   <warning>
+    <para>
+      While <command>CREATE INDEX CONCURRENTLY</command> or <command>REINDEX
+      CONCURRENTLY</command> is running on a unique index, <command>INSERT
+      ... ON CONFLICT</command> statements on the same table may unexpectedly
+      fail with a unique violation.
+    </para>
+   </warning>
+
   </refsect2>
  </refsect1>
 
diff --git a/src/backend/optimizer/util/plancat.c b/src/backend/optimizer/util/plancat.c
index 3b1ada1127e..47169e81d12 100644
--- a/src/backend/optimizer/util/plancat.c
+++ b/src/backend/optimizer/util/plancat.c
@@ -627,6 +627,11 @@ get_relation_foreign_keys(PlannerInfo *root, RelOptInfo *rel,
  * the purposes of inference.  If no opclass (or collation) is specified, then
  * all matching indexes (that may or may not match the default in terms of
  * each attribute opclass/collation) are used for inference.
+ *
+ * Note: during index CONCURRENTLY operations, different transactions may
+ * reference different sets of arbiter indexes. This can lead to false unique
+ * constraint violations that wouldn't occur during normal operations.  For
+ * more information, see insert.sgml.
  */
 List *
 infer_arbiter_indexes(PlannerInfo *root)