diff --git a/doc/source/testing.rst b/doc/source/testing.rst
index 659f23a..277f8a1 100644
--- a/doc/source/testing.rst
+++ b/doc/source/testing.rst
@@ -13,9 +13,8 @@ jobs.
 Unprivileged Single Use VMs
 ===========================
 
-All jobs currently run on these nodes. These are single use VMs
-booted in OpenStack clouds. You should start here unless you know you
-have a reason to use a privileged VM.
+All jobs currently run on one or more of these nodes. These are single
+use VMs booted in OpenStack clouds.
 
 Each single use VM has these attributes which you can count on:
 
@@ -33,22 +32,23 @@ Each single use VM has these attributes which you can count on:
   * It is also possible that these addresses are on multiple network
     interfaces.
 
-* CPUs are all running x86-64.
-* There is at least 8GB of system memory available.
+* CPUs are all running x86-64 unless you explicitly choose an AArch64
+  (64-bit ARM) label or nodeset.
+* There is at least 8GB of system memory available on default node
+  types, though we have limited availability of nodes with flavors
+  providing up to 32GB memory.
 * There is at least 80GB of disk available. This disk may not all be
   exposed in a single filesystem partition and so not all mounted at
   /. Any additional disk can be partitioned, formatted and mounted
-  by the root user; though if you need this it is recommended to use
-  devstack-gate which takes care of it automatically and mounts the
-  extra space on /opt early in its setup phase.
+  by the root user.
   To give you an idea of what this can look like most clouds just give
   us an 80GB or bigger /. One cloud gives us a 40GB / and 80GB /opt.
   Generally you will want to write large things to /opt to take
   advantage of available disk.
 * Swap is not guaranteed to be present. Some clouds give us swap and
-  others do not. Some tests (like devstack based tests) will create
-  swap either using a second disk device if available or by using a
-  file otherwise. Be aware you may need to create swap if you need it.
+  others do not. Some jobs will create swap either using a second
+  device if available or by using a file otherwise. Be aware you may
+  need add tasks to create swap within your job if you require it.
 * Filesystems are ext4. If you need other filesystems you can create
   them on files mounted via loop devices.
 * Package mirrors and/or caches for PyPi, NPM, Ubuntu, Debian, Fedora
@@ -78,8 +78,7 @@ Known Differences to Watch Out For
   even within the same cloud region.
 * Nested virt is not available in all clouds. And in clouds where it
   is enabled we have observed a higher rate of crashed test VMs when
-  using it. As a result we enforce qemu when running devstack and
-  may further restrict the use of nested virt.
+  using it. As a result we discourage general use of nested virt.
 * Some clouds give us multiple network interfaces, some only give
   us one. In the case of multiple network interfaces some clouds
   give all of them Internet routable addresses and some others do
@@ -87,9 +86,9 @@ Known Differences to Watch Out For
 * Geographic location is widely variable. We have instances all across
   North America and in Europe. This may affect network performance
   between instances and network resources geographically distant.
-* Some network protocols may be blocked in some clouds. Specfically
-  we have had problems with GRE. You can rely on TCP, UDP, and ICMP
-  being functional on all of our clouds.
+* Some Internet protocols may be blocked in some clouds. Specfically
+  we have had problems with GRE and multicast IP. You can rely on TCP,
+  UDP, and ICMP being functional on all of our clouds.
 * Network interface MTU of 1500 is not guaranteed. Some clouds give
   us smaller MTUs due to use of overlay networking. Test jobs
   should check interface MTUs and use an appropriate value for the