I suggest you delegate your rescue tasks to localhost - when things go wrong on remote hosts, the local host, where ansible is running, is the most likely thing to still be working. Documentation about delegation is here if you aren't familiar with it:
http://docs.ansible.com/ansible/latest/playbooks_delegation.htmlYou may need to adjust what you rescue actually does because of this - perhaps send a mail or message to get a human involved.
Hope this helps,
Jon