Introduce longer deployment wait with multi-node testing

The nature of multi-host interactions and testing can
be extremely racey. Since IPA can offer SSH, the wait
step can be short circuited resulting in us not waiting
long enough.

As such, introducing a longer wait _and_ re-check of SSH
connectivity.

While this does not completely fix the underlying race, it
is intended to help the test be more reliable in CI.

Partial-Bug: #1559764
Change-Id: Ief9ff9b2d158fa5322b5bb9be74105ba3d13b7fe

playbooks/roles/bifrost-prepare-for-test-dynamic/tasks/main.yml

@@ -29,7 +29,19 @@
   command: ssh-keygen -R "{{ ipv4_address }}"
   when: ipv4_address is defined
 - name: "Pause before asking for keyscan, to avoid races"
-  pause: minutes=2
+  pause: minutes=3
+  when: multinode_testing | bool == true
+- name: >
+    Re-check SSH connectivity prior to proceeding with multi-node testing
+  wait_for:
+    state: started
+    port: 22
+    host: "{{ ipv4_address }}"
+  when: ipv4_address is defined and multinode_testing | bool == true
+- name: >
+    Additional SSH startup pause when performing multi-node testing
+  pause:
+    seconds: "{{ node_ssh_pause }}"
   when: multinode_testing | bool == true
 - name: "Add testvm hosts from SSH known_hosts file."
   shell: ssh-keyscan "{{ ipv4_address }}" >> "{{ ansible_env.HOME }}/.ssh/known_hosts"