Currently /restart will restart Hudson immediately, even if jobs are still running.

In practice I usually /quietDown first, then update whatever plugins are new
etc., then /restart only when all currently running jobs have finished. This
ensures that jobs are not aborted unnecessarily, which can be relevant if you
are trying to match up test failures to changelogs.

But waiting for this to happen can take a couple of hours, so I have to be
careful to stick around watching the dashboard waiting for the executors to
clear. If I get distracted and forget to check for half an hour, that's up to
half an hour of time wasted when the server is idle.

It would be much friendlier if Hudson would just restart itself as soon as it
finished outstanding jobs.

Specific suggestion: if /restart or /exit is invoked,

1. Do the equivalent of /quietDown immediately;

2. Start listening for all executors to be idle, and actually restart or exit
when that happens;

3. Display a response page containing a button to restart or exit immediately.

Should not be hard to implement, though I am not sure about #2; I can't find any
global listener which would be notified on all Executor events (on master as
well as slaves).