When I use Whirr, I compose whirr.config (common stuff + variable stuff + derived stuff) where derived stuff is calculated from variable stuff like type of machine.
During post-processing when Whirr finishes, I "cp -r" the directory Whirr created and copy the composed whirr.config into it so I can keep notes about a run.
It would be nice to have this built-in ("keep instead of remove after terminated"). I also have a snapshot mechanism that captures a tar-gz of a list of files relative to / to capture config files "as built". That way I can examine them after the ensemble is terminated.
As mentioned in another jira item the instances file generated by Whirr can be used for reference and to avoid expensive ec2 api (or other provider equiv) calls.
Log output of Whirr is best kept in the directory it generates.
Whirr cluster destroy can touch a file "_Terminated" after termination.