Implement a trap for SIGUSR2 that will immediately wake a sleeping chef-client and trigger a run, without waiting for splay.
Currently, a daemonized chef-client will respond to SIGUSR1 by waking up the sleeping process. Chef-client, however, will still respect the random splay configured by the user.
While perhaps unintentional, this behavior is useful if you want to run chef-client on a large number of nodes, do not want to wait for the normal daemonized run, but still want to avoid a bit of the Thundering Herd.
But, sometimes I want to run Chef immediately, without any splay.