Jobs timing out when trying to ssh at cleanup stage

I've seen several hung jobs with this sort of error where it appears the ssh connection fails. I've seen it on

http://build.ros.org:8080/view/Queue/job/Kbin_dj_dJ64__grid_map_visualization__debian_jessie_amd64__binary/36/console

00:11:23.242 SSH: Connecting from host [ip-172-31-7-50]
00:11:23.245 SSH: Connecting with configuration [repo] ...
02:00:01.258 Build timed out (after 120 minutes). Marking the build as failed.
02:00:01.263 ERROR: null
02:00:01.263 Build step 'Send files or execute commands over SSH' changed build result to FAILURE
02:00:01.263 Build step 'Send files or execute commands over SSH' marked build as failure
02:00:01.264 $ ssh-agent -k
02:00:01.265 SSH: Caught exception [Failed to read file - filename [/var/lib/jenkins/.ssh/id_rsa] (relative to JENKINS_HOME if not absolute). Message: [java.lang.InterruptedException]] Sleeping for [5,000]ms before trying again
02:00:01.276 unset SSH_AUTH_SOCK;
02:00:01.276 unset SSH_AGENT_PID;
02:00:01.276 echo Agent pid 2392 killed;

http://build.ros.org:8080/view/Queue/job/Kbin_uX32__grid_map_octomap__ubuntu_xenial_i386__binary/1/console http://build.ros.org:8080/view/Queue/job/Kbin_uX32__grid_map_visualization__ubuntu_xenial_i386__binary/35/console http://build.ros.org:8080/view/Queue/job/Kbin_dj_dJ64__grid_map_rviz_plugin__debian_jessie_amd64__binary/40/console

Potentially all builds on this computer? http://build.ros.org:8080/computer/ip-172-31-7-50.us-west-1.compute.internal/

http://build.ros.org:8080/computer/ip-172-31-7-50.us-west-1.compute.internal/builds

ros-infrastructure / ros_buildfarm

Jobs timing out when trying to ssh at cleanup stage #466