The steps involved in startup/shutdown and the conditions that might cause them to take longer/shorter are not well understood. We should produce a list of all known startup/shutdown tasks, along with the following information:

• Owning team
• Success/Failure cases, if any
• Any conditions that affect how long this task might take.

Ideally, this list would be kept up to date, possibly by being checked into the server codebase as an architecture guide.