After we restart a `repo` host, all the repository daemons jump into their full fetch/pull cycle from the beginning. This can create high amounts of load on the box:
> load average: 81.65, 34.03, 14.01
This only lasts for a minute or two, but it might be nice to do something like:
- Store update state on the Repository objects so restarting doesn't re-pull everything immediately (there are plenty of ways to prioritize an update already).
- Have the script examine repository state on startup and immediately delay pulls according to last commit time. This seems simpler than the above and probably better in every way.
- Have some kind of "warm start" flag which activates the above behavior.