Exercise 3: A Checklist for Long Jobs on Remote Servers — Possible Solution
=============================================================================

Before starting
  [ ] Is screen installed? (screen -v)
  [ ] Is this really a one-off job? If it should run at boot or restart
      when it fails, make it a systemd service instead.

Starting
  [ ] Start a named session, with logging:   screen -S name -L
  [ ] Name the window (Ctrl-a A) and redirect or log important output
  [ ] Turn on silence monitoring (Ctrl-a _) to hear when it finishes

Checking progress
  [ ] Detach with Ctrl-a d — never type exit
  [ ] From outside: screen -ls, tail screenlog.0, or
      screen -S name -X hardcopy /tmp/snap.txt

Surviving a disconnection
  [ ] Reconnect and run screen -ls
  [ ] Detached -> screen -r name;  Attached -> screen -d -r name
  [ ] If sessions keep vanishing at logout, check KillUserProcesses

If I forgot to use screen
  [ ] Ctrl-Z, bg, then reptyr PID inside a screen window
  [ ] No reptyr? Ctrl-Z, bg, disown %1 (and accept I can't watch it)

Cleaning up
  [ ] Confirm the job finished successfully (scroll back or read the log)
  [ ] screen -S name -X quit, then screen -wipe
  [ ] Delete old screenlog files you no longer need

WHY THIS WORKS AS AN ANSWER
------------------------------
It turns the course into a routine you can follow without thinking,
including what to do when something has already gone wrong. Your own
list may differ; what matters is covering all four stages.