[JabavuAdams]

Even if that's true, consider how long it takes to type the command to reboot a server. Let's say pg reboots it 70 times a year. So, maybe he spends 1 hour to 1.5 hours a year rebooting HN. That's a better use of time than spending 2+ hours to make the site more scalable. There's no one he's interested in (i.e. deal flow) who's going to stop using HN or applying to YC because the site goes down every 5-6 days.

There may be all kinds of other reasons to update the code, but that comes down to available time, experimentation, and enthusiasm, not productivity.

[Terretta]

And could run it under an angel, auto rebooting when mem leak above a certain threshold and traffic below a certain level, allowing a couple days' window to look for a low impact time.

This is more or less what Apache and IIS do with child processes or long running threads. It's a production proven approach.

[peterwoo]

I wonder what (# of expired links clicked each year * 8 seconds / link) amounts to.