TODO

Go through a full run of test data, printing out exactly what it does, and find all the problems. Also, view Discourse after a full run and see all the problems.

Figure out what the "final composite" score should look like. Right now, just looking at execution time for threads.

Figure out how to handle Bundler.with_clean_env for start script invoking Discourse. Right now it's a mess, and you can't "bundle exec ./start.rb", which isn't great.

Add environment vars to allow easily simultaneous building of multiple AMIs with the same EC2 creds, on the same build laptop