Test cases

• Task fails to start: change one task's port to match the other
• Task errors at runtime: HTTP GET :8081/task1/quit
• HTTP server shutdown errors: change if false to if true in the case <-stopCh: branch of a task
  • To see it all work together, do this and then hit /quit on the other task
• Emergency shutdown: hit ^C twice in quick succession

Philosophy

This is classic programming-in-the-large. Other approaches would be to

• Have the tasks isolated within the process somehow, and rather than report their errors, Let It Crash (see Erlang) and restart. Go's errors-as-values makes this tricky but it can be done.
• Build the tasks into separate binaries and run in separate unix processes, then defer to something like Kubernetes as the supervisor, again Erlang-style.