Clarify (lack of) backup redistribution on data structure level [HZG-…

…156] (#1313)

Adds an info-box to clarify the following sentence.
> The distribution happens on the partition level; the data and its backups are stored in the memory partitions.

docs/modules/fault-tolerance/pages/backups.adoc

@@ -7,6 +7,15 @@ The distribution happens on the partition level; the data and its backups are st
 memory partitions. See the xref:overview:data-partitioning.adoc[Data Partitioning] and xref:clusters:partition-group-configuration.adoc[Partition Grouping]
 for more information about the partitioning.
 
+[IMPORTANT]
+====
+Hazelcast does not support backup redistribution on a data structure level. Thus, changing the backup count of a nonempty data structure is **not** supported. For example, for a map:
+
+1. increasing the backup countfootnote:change-backup-count[The backup count of an empty map can only be changed by modifying the xref:configuration:understanding-configuration.adoc#static-configuration[static configuration] and restarting the cluster.] does **not** create additional copies of existing map entries,
+2. decreasing the backup countfootnote:change-backup-count[] does **not** remove any backups of existing map entries, so some backups will have stale data if the map entries are overwritten, and
+3. if some backups are lost due to node failure, additional copies are **not** created on other nodes to meet the backup count.
+====
+
 When a member in your cluster is lost, Hazelcast redistributes the backups
 on the remaining members so that every partition has a backup.
 The number of backups is configurable.