[backport-1.4]: need to improve how cluster-machines-ready Job times out

A solution for issue #2216 (closed) needs to be backported to 1.4.x