I appears that some buildslaves are left running after cp has flagged a tegra as offline and stops monitoring. Two things need to happen - find out why and add a check during monitor loop to detect if/when this has happened
During the mtv outage, this is causing staging to burn, but production seems ok ? I wonder what the difference is, other than master buildbot versions.
this was fixed when cp was refactored for bug 610600 and the killPID loop was cleaned up