Were we nervous, or what?

There are always surprises lurking out there. Just to complicate things, during production rollout we implemented not only the Smalltalk image, but also a few 100K of code changes on both sides of the system

Having development tools in the production image leads to temptation. Only one post implementation change was made to the production image in violation of testing practices. This is a good record, considering the temptation just to fix things on the fly.

Don't forget to secure access to the app, and later someday deploy it without the compiler. Don't forget to fully test the stripped image (its amazing what you might cut out...)

Queue counts and flashing lights are great for general feelings, but can we do better? Remember, server performs repetitive functions. Lets look at an EKG, the heartbeat of the application.