/usr/bin/ruby # wraps system installed rbenv (available during ubuntu boot)

ruby.rb

#! /bin/sh

# file: /usr/bin/ruby
#
# the normal 'rbenv global 1.9.3-p0' will affect this script but, unlike having *only* rbenv, this ruby will be available at low run levels (aka, for init scripts)

export RBENV_ROOT="/usr/local/rbenv"

export PATH="/usr/local/rbenv/bin:/usr/local/rbenv/shims:$PATH"

exec ruby "$@"

Interesting, unfortunately, or fortunately - depending on pov [worker or mgt :)], our policy is a "only do it once by hand". Pretty much once we have some thing or work-flow working, it has to be converted to a script. So we do "roll by hand", but that thing or work-flow has to survive the removal of the person ;)
Essentially, in our case production means that person-less version. E.g Out the box logging is then quite important so that we can easily grep/parse for any issues. But in your circumstances it looks like a total waste, and it is in those circumstances.
Of course, like your EBS, we only build one AMI's perregion, once, then they automatically get cloned to a new AMI if they build successfully - all fully automated of course ;)

Appreciate any further thoughts you have that might improve things.