Improve allocation-disabling instructions

Clarify the “one minute” in the instructions to disable the shard allocation
when doing maintenance to say that it is configurable.

Add a note about making sure that no rebalancing occurs until the maintenance
is complete.

Relates elastic#19739.

docs/reference/upgrade/disable-shard-alloc.asciidoc

@@ -1,8 +1,10 @@
 
-When you shut down a node, the allocation process waits for one minute
-before starting to replicate the shards on that node to other nodes
-in the cluster, causing a lot of wasted I/O. You can avoid racing the clock
-by disabling allocation before shutting down the node:
+When you shut down a node, the allocation process waits for
+`index.unassigned.node_left.delayed_timeout` (by default, one minute) before
+starting to replicate the shards on that node to other nodes in the cluster,
+which can involve a lot of I/O.  Since the node is shortly going to be
+restarted, this I/O is unnecessary. You can avoid racing the clock by disabling
+allocation before shutting down the node:
 
 [source,js]
 --------------------------------------------------
@@ -14,4 +16,14 @@ PUT _cluster/settings
 }
 --------------------------------------------------
 // CONSOLE
-// TEST[skip:indexes don't assign]
+// TEST[skip:indexes don't assign]
+
+Additionally, the allocation process attempts to keep the shards balanced
+across the cluster by performing rebalancing, moving shards between nodes. This
+involves I/O that is also unnecessary if the shut-down node will shortly
+return, so it is preferable for the cluster not to attempt to perform any
+rebalancing until the maintenance is complete. This is the default behaviour,
+but can be changed using the `cluster.routing.allocation.allow_rebalance`
+setting. Make sure that this setting is unset, or set to the default value of
+`indices_all_active`, before continuing.
+